The Design and Realization of CoViD, A System for Collaborative

The Design and Realization of CoViD, A System for Collaborative
The Design and Realization of CoViD,
A System for Collaborative Virtual 3D Design
Wolfgang Stuerzlinger, Loutfouz Zaman, Andriy Pavlovych
York University
Toronto, Canada
Ji-Young Oh
Univ. of Arizona
Tucson, AZ, USA
http://www.cs.yorku.ca/{~wolfgang|~zaman|~andriyp}
jyoh@optics.arizona.edu
signer(s) have proposed a first version of the design. There,
typically in a meeting, the stakeholders in the project get
together and evaluate these potential solutions. Frequently,
the participants in this meeting want to interactively modify
the proposed designs to explore the design space better.
Today's design tools and computer infrastructure do not
support such activities well, as computer systems typically
support only a single user and computer-aided design tools
require significant training.
ABSTRACT
Many important decisions in the design process are made
during fairly early on, after designers have presented initial
concepts. In many domains, these concepts are already realized as 3D digital models. Then, in a meeting, the
stakeholders for the project get together and evaluate these
potential solutions. Frequently, the participants in this meeting want to interactively modify the proposed 3D designs to
explore the design space better. Today's systems and tools
do not support this, as computer systems typically support
only a single user and computer-aided design tools require
significant training.
Traditional tools for 3D design require a large amount of
training. Part of this is based on the fact that they offer a
large number of functions to support every possible 3D
design activity. Another issue is that traditional 3D design
tools expose the technical foundations of computer graphics
directly to the user. The problem is that, without training,
most people cannot easily understand the visualization of
3D objects in multiple orthogonal views or do not understand the intricacies of hierarchical 3D transformations.
This paper presents the design of a new system to facilitate
a collaborative 3D design process. First, we discuss a set of
guidelines which have been introduced by others and that
are relevant to collaborative 3D design systems. Then, we
introduce the new system, which consists of two main parts.
The first part is an easy-to-use conceptual 3D design tool
that can be used productively even by naive users. The tool
provides novel interaction techniques that support important
properties of conceptual design. The user interface is nonobtrusive, easy-to-learn, and supports rapid creation and
modification of 3D models. The second part is a novel infrastructure for collaborative work, which offers an interactive table and several large interactive displays in a semiimmersive setup. It is designed to support multiple users
working together. This infrastructure also includes novel
pointing devices that work both as a stylus and a remote
pointing device. The combination of the (modified) design
tool with the collaborative infrastructure forms a new platform for collaborative virtual 3D design. Then, we present
an evaluation of the system against the guidelines for collaborative 3D design. Finally, we present results of a preliminary user study, which asked naive users to collaborate
in a 3D design task on the new system.
One common solution to offer multiple users access to a
computer system is to use one (or more) large display, typically via projection. However, most large display systems
allow for only one active user at any given time, even if
multiple physical screens are available! The reason behind
this is that most software packages and graphical user interface toolkits can only handle input from a single user. This
leads to the “driver” problem, i.e. that one person controls
the content and collaboration of the meeting – usually the
person who is most adept in controlling the system, which
may not be the person best suited to edit a design.
This paper introduces a new system that aims to make the
collaborative 3D design process more productive. The factors mentioned above currently render 3D design a suboptimal and time-consuming process. To illustrate the potential gains, we present the following scenario: Imagine a
family who wishes to redesign their living room with the
help of a professional interior designer. Traditionally, and
after an initial consultation, this designer came up with several concept designs and presented them to the family. The
family picks one (or two) designs and proposes some modifications to adapt the design to their needs. Several days
later, they meet again over a more refined version of the
designs, which addresses the needs of the family better. As
the designer may have not have addressed all concerns fully,
this process is usually iterated a few times. Also, in this
Keywords: collaborative design, 3D design, collaborative
virtual reality
INTRODUCTION
Today, digital 3D models are critical in many domains,
such as architecture and urban planning, all kinds of industrial design, the entertainment industry, and many engineering applications. Many of the important decisions surrounding a design are made in the initial phases, after the de-
1
process, auxiliary information, such as furniture catalogues
and images of other living rooms are consulted. At last, a
final design is chosen and the designer elaborates it further
so that other people can build the desired living room configuration.
naïve users to quickly create 3D designs. SESAME is one
of the main components of CoViD and will be discussed in
detail later.
Collaborative systems have been studied in the computer
supported collaborative work (CSCW) and groupware literature for many years. While distributed collaborative systems are available, we point out that face-to-face interaction
is very valuable for collaboration. This is especially true for
the decision making-process as people are usually reluctant
to negotiate important issues over the distance. Hence, this
system focuses on collaboration between people in a single
location in a form of meeting. As such meetings typically
take place around a table, we focus on systems that include
some form of interactive tabletop system.
With a collaborative 3D design system, this process could
work as follows: After the initial consultation, the designer
brings digital versions of several concept designs to the
meeting with the family. All participants sit around a fully
interactive system, which allows each of them to directly
modify the digital model of the initial concepts. While the
technical system may support fully simultaneous operations,
the normal social protocols encourage the members of the
family to take turns and to work constructively with the
designer. Auxiliary information is displayed on a secondary
screen, typically by browsing an on-line catalogue or digital
images. Based on the direct visualization of the 3D design,
the family and the designer quickly agree on one alternative.
Then the designer again elaborates on the final 3D design
and passes it on to others.
Standard computer systems are designed to support interaction with only one user at a time. If more users want to use
the system, they must take turns. This limitation motivated
researchers to come up with various kinds of groupware
systems allowing multiple users to perform tasks simultaneously on one (or more) display. This is commonly referred to as Single Display Groupware [14] or Shared Display Groupware (SDG). The main issues here are that SDG
hardware needs to support multiple input devices and that
the software running on the SDG hardware needs to support
(potentially simultaneous) input from multiple users correctly. An overview over recent work in tabletop SDG systems can be found in Scott et al.’s review of the area [13].
As this example highlights, the possibility to quickly manipulate a 3D digital model in a collaborative setting enables a much more rapid design process. This paper presents a system that targets this scenario.
RELATED WORK
Due to space restrictions we cannot give here a complete
overview of all work in all relevant areas. Hence, we point
to publications that provide good overviews of related work
and discuss only the most relevant pieces of work.
Finally, there has been some previous work in collaborative
3D design. One of the earliest was the Teledesign system
[15], but there each user sat in front of their own display. A
collaborative SDG 2D design system had been previously
been explored by Tse [17] and Kidpad explored a collaborative environment for kids, where multiple children could
simultaneously draw on one display [2].
There are many approaches to 3D design. Most of them can
be categorized into traditional 3D computer-aided design
systems, sketch reconstruction systems, gesture-based systems, voxel-based systems, virtual reality systems, and systems that are targeted at naïve users. A recent overview of
most of these areas can be found in [8].
Guidelines for Collaborative Design Systems
Previous work in 3D design as well as collaborative systems has resulted in a set of guidelines that enumerate desirable criteria for such systems. The first part of this section lists a set of guidelines for the design process, which
are based on a review of the design research literature [7],
which also contains more details.
As the computer skills of participants in collaborative work
often varies substantially, most systems that require that the
user has a good understanding of 3D geometry cannot be
used. Similarly, systems that suffer from various limitations
of today’s recognition, 3D display, or 3D tracking technologies are also a bad choice as user frustration is quickly
fatal for collaboration. This leaves only easy-to-use 3D design systems as a viable choice for collaborative design.
• Non-intrusive interface: One of the reasons that designers
sketch is due to the complexity of design problems and
the fact that mental resources are limited. Therefore, designers sketch to externalize their vague ideas in their
mental imagery and to visually evaluate them. Hence, the
cognitive demand for using computer tools should be
minimal so that the designers can fully commit themselves to solve the design problem rather than worrying
about the tool interface.
SKETCH [18] is one of the most prominent examples, but
it still necessitates the user to learn a predefined set of gestures. Teddy [5] is a system limited to the design of freeform humanoid and animalistic shapes, but it has been
shown to be usable even by children. Sketchup by @Last
Software (www.sketchup.com) is an effective commercial
tool for 3D design, which uses extrusion as the main way to
create 3D objects. Another system that was developed at the
same times and shares some of it’s features is SESAME [7,
8, 9, 10] (Sketch, Extrude, Sculpt, and Manipulate Easily),
a solid-modeling conceptual design tool that allows even
• Easy creation: The design process is dialectic and cyclic.
Solutions are repetitively created or refined while (mentally) testing them against the desired criteria. This means
that the cost to visualize a solution should be minimal.
2
Therefore, any computer tool has to provide efficient
ways to create design solutions quickly.
guidelines for tabletop displays, which is summarized here.
For details and references please consult the original article:
• Easy combination and restructuring: The recognition of
individual parts and their relationships aid a human in the
interpretation of a whole object. In the context of design,
researchers observed that the combination and restructuring of parts are the main activities in creative invention.
Furthermore, combination is a simple mental activity,
while restructuring requires the aid of externalizations.
Therefore, the user must be able to combine components
of objects easily as well as to be able to restructure the
result of a combination.
• Natural interpersonal interaction: Collaboration works
only when people can communicate effectively. Any
breakdown in communication (be it due to technical
problems, technical limitations, etc.) has strong adverse
effects on the ability of people to achieve work together.
• Transitions between activities: Many tasks require that
users can seamlessly and quickly switch between various
activities. For example, it may be necessary to switch between various drawing modes, keyboard entry, different
software tools (e.g. design tool, WWW browser, spreadsheet), and different forms of collaboration (presentation,
collaborative work, decision making, etc.).
• Tolerance to ambiguity and incompleteness: Although
design decisions are not well formed in the conceptual
design phase, designers can still express their ideas via
sketching, and the visuals naturally exhibit the ambiguity
and incompleteness of these ideas. Hence, computer tools
should not always expect precise input from users, and
provide a way to externalize ambiguous forms. In addition, the visual output has to reflect the tentativeness of a
solution, so that the designers can easily identify newly
created problems or defects from intermediate forms.
• Transitions between personal and group work: In collaborative work, people transition fluently between individual and group work. The support of a “personal space”
for every participant is important to facilitate experimentation. This can be accomplished via external devices (i.e.
a portable computer for each person) or by “partitioning”
a large display surface appropriately.
• Transition between tabletop collaboration and external
work: Collaborative work often involves the integration
of independently developed pieces into the larger work as
well as the reverse. Furthermore, it needs to be simple to
transfer content that was externally developed or modified into a system and out of it.
• Range of levels of abstraction: There is a range of levels
of abstraction that designers commonly move within,
since they can only deal with a limited set of problems at
any instant. Experienced designers tend to shift more fluently between overall and detailed aspects of design. This
range may vary between different design disciplines (architect vs. door designer) and any computer tool has to
match to the range of detail that a designer works with.
• The use of physical objects: Table surfaces provide a
convenient location to place objects such as laptops,
notebooks, printouts, and even non-task-related objects
(cell phones, etc.). Any good tabletop infrastructure
should not block this kind of usage. Furthermore, the idea
of tangible user interfaces has been introduced. Here special objects tracked by the infrastructure are used as input
devices (e.g. the use of a model house to specify the location of a building in the design).
• Ability to edit various forms of information: The representations used in the design process are not only geometric shapes, but also different free-form strokes that
stand for size, ratio, or trajectory. By putting figural and
conceptual information together, a designer can reflect on
different dimensions of a design problem at once. Therefore, the goal of sketching is to organize the problem/solution via different kinds of symbolic representations in the course of producing a final geometric shape.
Many tools overlook this factor by focusing overly on
various geometric representations.
• Accessing shared physical and digital objects: As participants are typically placed around a table in a meeting,
shared access to the artifacts becomes a necessity. For
physical objects, this is less of a problem as people are
used to sharing objects on a table surface and can easily
decode pointing gestures towards them. For shared digital
objects this can be more of a problem, when objects are
duplicated for each user, and it is usually simpler to have
a single visualization accessible to everyone. Furthermore, the orientation of each object may pose problems
on a tabletop surface – e.g. it is hard to read and interact
with upside down text. Finally, a user of the infrastructure should not be able to block the display for another
user (e.g. with the shadow of their hand).
• Supporting evaluation (simulation): Designers explore a
solution/problem space by generating many solutions and
testing them, asking ‘what if’. Sketching visualizes a
situation on paper, and designers perform simulation of
the situation in their mind. On the other hand, computer
tools can conduct the simulation directly, instead of the
designers. This should be beneficial, if the system can
support spontaneous creation, modification, and resimulation. However, simulation is closely task-specific,
so one cannot count this as a criterion to judge computer
tools for the design process.
• Flexible user arrangements: Users may want to sit or
stand in many configurations around a table, depending
on the type of collaboration (decision making, presentation, collaborative work, …), the interpersonal relations
In the second part of this section we list a set of guidelines
for collaborative work. Scott et al. [13] created a list of
3
between participants, and the properties of the task at
hand. A tabletop infrastructure should support all these
and allow for fluent transitions between various forms of
arrangements.
high level, the tool has two main modes: 2D and 3D, each
of which is activated once any tool from the corresponding
palette is selected. However, there are several operations
that are available in both modes. An overview of the mouse
function assignments is shown in Table 1.
• Simultaneous user interactions: When multiple people
engage in tabletop activities, they often and naturally interact with artifacts on the table surface simultaneously.
Many traditional computing platforms support only one
input device and hence force users to take turns. This
leads to a style of interaction where a single person
“drives” at a time, which is hinders fluid collaboration.
Supporting multiple users simultaneously is both a hardware and a software issue. For the hardware side, the infrastructure needs to be able to reliably track and distinguish between individual interaction devices with imperceptible latency. For the software side, the infrastructure
needs to support multiple concurrent actions (potentially
even with different applications simultaneously).
Input command
2D
Function
Select 2D shape
Select or add 3D
object
Draw 2D shape
3D
Move 3D object
2D
Clone 2D shape
Clone 3D
object
Rotate
3D object
Mode
2D
Left click
3D
Left drag
Shift +
left drag
3D
Ctrl +
left drag
3D
Middle button
COVID – A SYSTEM FOR COLLABORATIVE VIRTUAL
3D DESIGN
The CoViD (COllaborative Virtual 3D Design) system consists of two main parts – the SESAME conceptual design
tool and the MULTI collaborative setup. To simplify the
discussion, we first discuss the single-user version of
SESAME, then the MULTI platform, and finally how both
components work together in CoViD.
Both
Realization
On release
On press
On dragging more
than a threshold
distance
On dragging more
than a threshold
distance
During dragging
Navigation
Right drag
towards outside Both
of a volume
Extrude a 2D
contour or a face
During dragging
Right drag
towards inside
of a volume
Both
Subtract an
extruded shape
from a 2D contour
or a face
Extrude during
dragging, subtract
after release
Shift +
right drag
Both
Scale object
During dragging
Table 1. Mouse and keyboard commands.
SESAME - Sketch, Extrude, Sculpt, And Manipulate Easily
Cloning provides a powerful way to create repetitive patterns in combination with the grouping technique explained
below. It is implemented as a (continuous) dragging action.
That is, once a user drags the selected source object more
than certain distance, a cloned object is instantiated. The
cloned object will continue to move in the scene, until the
user places it on the target position by releasing the mouse.
SESAME was designed to be a simple-to-use 3D design
tool for 3D design. It is an ideal tool in the context of
CoViD, as in any collaborative setting the computer skills
of the participants will vary wildly.
The user interface of
(the
single-user
version of) SESAME
consists of a main 3D
scene view and a menu
panel on the right side.
The menu offers a
color/texture
palette
(top right), a 2D and
3D primitive shape
selection palette (left
part of menu panel), an Figure 1. SESAME user interface.
undo button (at the
bottom of the 3D palette), a navigation mode switch button,
and a recycle bin (Figure 1). The 2D palette consists of
common 2D tools such as lines and arcs, and a freeform
drawing tool. The 3D palette provides a tool to move objects in 3D and the instantiation of primitive 3D shapes,
such as boxes, triangular prisms, spheres, and cylinders.
Additionally, SESAME provides a simple navigation interface so that users can assess the 3D structure of the scene
rapidly. Camera rotation, pan, and zoom are accessible
through middle mouse button dragging, shift-dragging, and
scrolling, respectively.
Improved Tools for 3D Design
Beyond the basic actions listed above, SESAME provides
also several advanced facilities for the rapid creation of 3D
content.
With the 2D drawing tools the user can draw lines, arcs, or
free-form curves onto any planar surface. As the user draws
these in a 3D perspective view, SESAME displays a perspectively distorted circle during 2D drawing operations to
help the user perceive the orientation of the current drawing
plane. That is, in addition to visualizing the current line
with the well-known rubber band technique, a circle is displayed with the origin at the start of the line and a radius
proportional to the current length of the line (see Figure 2).
The user interface utilizes only a 3-button mouse and a few
modifier keys. All actions such as sketching a contour or
moving objects can be accomplished with them. From a
4
Figure 2. A circle is used to visualize the orientation of the
current plane in SESAME. The user is currently drawing line
a. Additionally, multiple colored guides are displayed based on
other connected lines (b and c). The yellow line is perpendicular to b, the purple line is parallel to c, and the straight white
lines are parallel to the coordinate axes.
(a)
(b)
(c)
(d)
(e)
(f)
Figure 3. Sketching interface. (a-b) Dragging outwards from a
closed contour creates a volume. (c-d) Drawing a contour on a
face and dragging inwards sculpts the volume. (d-e) The user
creates a new shape by dragging outward. (e-f) The user can
also “stretch” an object by dragging the face directly. (Red
lines added to visualize drag operations)
Furthermore, multiple guides are displayed as well. These
include guides parallel to the coordinate axes, as well as
suggestions for lines that are perpendicular and parallel to
other parts of the 2D drawing. The cursor snaps to these
suggestions to aid the construction of common configurations. To avoid overloading the user, the SESAME displays
only suggested completions that are close to the cursor.
Finally, SESAME also provides support for the selection of
groups of objects. For this, the tool analyzes the scene and
detects which objects are placed on top of another, as such
objects will move with the base object in the real world.
The implementation then simply selects objects on top of a
base object whenever the base is selected (e.g. with clicking
or rectangle selection). This turns out to greatly facilitate
the manipulation of common scene configurations [9].
Lastly, a multi-click scheme, similar to character, word and
paragraph selection in MS Word, cycles through an object
with all objects it supports, the group consisting of all
touching objects, and the object itself. Together, these selection techniques make restructuring and combinations of
many 3D scenes easier.
Whenever the user adds any 2D shape, SESAME analyzes
the current drawing and detects all closed contours created
by the added shape. This also allows the user to subdivide
existing closed contours by drawing other shapes over it.
This effectively accommodates the creation of emergent
shapes as these have been shown to facilitate creative thinking during the design process [3, 4]. The user can then extrude any closed contour by clicking with the right mouse
button inside the contour and dragging. For this, the height
of the extrusion is proportional to the length of the drag, i.e.
the top surface of the extrusion follows the cursor. A drag
operation outwards relative to existing 3D geometry will
extrude a new volume (or resize an existing volume), while
a drag inwards will sculpt the volume, i.e. can be used to
create holes. Figure 3 illustrates the technique.
Evaluation of SESAME
Traditional computer design tools are targeted towards the
final stages of design. Senior designers have commented,
“recent designs are very beautiful due to the [use of a computer] design tool, but they are very poor, far from the design ideal” [16]. This means that while all details are perfect, the users of current design tools often do not adequately explore the space of all possible design solutions.
SESAME also provides a technique to let the user grab any
object and slide it across the scene to the desired position to
visually assess the impact of a change. One of the main
motivations behind this is that this will greatly facilitate
exploration. As almost all objects in the real world are attached to other objects, we based the design of our interaction technique on the idea that, unless special actions are
taken, objects should always stay connected with another
part of the scene and should not interpenetrate other objects.
This conforms better to the way most people think about the
real world. In contrast to other techniques, SESAME includes a novel technique that uses the entire area of the visual overlap of a foreground object with the (potentially
complex) background, as this has been shown to work very
well for 3D object manipulation [6]. An evaluation showed
that this new technique is significantly more efficient for
novice users and showed that our technique conforms very
well to users’ expectation about the position of objects relative to a scene.
SESAME was previously evaluated to analyze its suitability
as a design tool. A comparison with sketching with pen and
paper revealed that SESAME yields the same level of design quality as judged by an expert designer [7]. Another
comparison with users of 3D Studio Max demonstrated that
even first-time users of SESAME could generate meaningful designs faster with this tool than traditional CAD systems [10]. Also, studies with naïve users (i.e. people who
had never used 3D design systems before) showed that
most people quickly understand how SESAME works and
are rapidly able to use it to create 3D designs.
SESAME, as described, is capable of creating mainly prismatic objects, i.e. everything that can be modeled by extrusions. This limits the application domain to engineering
domains, interior design, as well as architecture. Current
5
work on SESAME aims to add support for rotationally
symmetric objects as well as freeform modeling tools.
cameras are placed behind the screens (both for the table
and the walls).
Each of the three back-projection screens has a 60” diagonal and is illuminated by a NEC WT600 projector, which
have an extremely short throw distance. This helps to reduce the space usage of the configuration. The tabletop has
also a 60” diagonal and two WT600 projectors illuminate it
from underneath (the front and back half of the table, respectively). This “split-screen” design gives users free access to the two long sides and one short side of the table.
Careful positioning of the projectors guarantees an almost
seamless image. All projection screens are acrylic with a
0.7 gain. A pane of tempered glass thick enough to carry
the weight of an adult supports the tabletop screen and a
thin protective layer on top prevents scratches. Finally,
there is a 15 cm wide metal ledge around the three accessible edges of the tabletop.
MULTI – Multi-User Laser Table Interface
In general, the stakeholders in a design project make important high-level decisions in a meeting. However, for effective collaboration, the participants in such a meeting need to
be able to see a visualization of the current design proposal(s). If this visualization is interactive, the participants
will in general be able to reach better decisions, as they can
explore the design space more completely.
However, 3D design activities frequently require large displays, as they make it simpler to assess the impact of a proposed design. Another factor is that screen space is often
limited in many situations, especially when multiple proposals are evaluated simultaneously. Furthermore, in collaborative meetings, there is always secondary information
that is critical to the task, but not represented on the main
visual display. Often this information is available in printouts, but this is clearly not a very interactive medium. Another alternative is various kinds other mobile devices, but
all of these are essentially single-user devices and are often
poor platforms for collaborative activities. As discussed
above, most meetings take place around tables. However,
[12] stated that for groups with more than three persons a
single tabletop is not sufficient and that larger groups may
well need vertical displays for shared information. Hence, it
is desirable to have an infrastructure that supports both an
interactive tabletop as well a several additional large interactive display surfaces.
The MULTI (Multi-User Laser Table Interface) infrastructure was designed for collaborative 3D design. It consists of
an interactive table and three interactive walls that are positioned in a semi-circular arrangement around one of the
short sides of the table. A single 3 GHz Windows XP computer, with three graphics cards with two outputs each,
drives all 5 projectors. The infrastructure includes also a
standard 802.11 wireless network and several Tablet PC’s
(not shown). An image of MULTI is shown if Figure 4.
Figure 4. An overview of the screen configuration of MULTI.
All projectors and cameras are located behind the screens.
The image on the table is created by two projectors (partially
located underneath the middle wall screen).
The main interaction devices of MULTI are several computer-controlled laser styli. The laser spots observable at
any intersection of the laser beam with the projection surfaces are detected by a set of cameras. Each stylus works
both as a pen (as in pen-based computing) as well as a remote pointing device. The button on a stylus is configured
to work as the left mouse button, which makes it very natural to interact with standard GUI applications. Each laser
stylus is wired to a control box that allows the computer to
control their laser diode and the status of the button on the
stylus. The laser diodes are driven with a time-multiplexing
scheme that allows the infrastructure to identify each laser
stylus [11]. In the MULTI system, we use a “scaled version” of this time-multiplexing scheme that allows simultaneous tracking of 7 laser styli on all 4 surfaces. All laser
spots are tracked with Naturalpoint cameras, which feature
partial hardware support for the detection of bright spots.
We removed the IR filter from each camera to enable tracking of the laser spots via the standard CMOS camera chip.
MULTI is designed for groups of 5 to 7 active participants
and the size of the table was chosen so that five people can
comfortably sit around it. To give users full access to both
long sides and one of the short sides of the table, we had to
use two projectors for the table, each of which covers approximately half of the table. The decision to add several
walls was motivated by several factors: the desire to be able
to work on large-scale designs, the desire to support even
more users as a “passive” audience, and the ability to use
e.g. only the walls alone for large-scale visualization. Furthermore, the table has a stable ledge around the interactive
surface in the middle, which provides space for laptops,
paper, and other work artifacts (and is even stable enough
to sit on).
To avoid issues with shadows cast by user’s hands etc., all
surfaces are back-projection screens and the projectors and
6
Each camera has a 57º lens, but due to sightline restrictions
MULTI uses 6 cameras underneath the table, 4 behind the
middle wall, and one behind each sidewall. As the time
multiplexing of the laser diodes is synchronized with the
refresh rate of the cameras (120 Hz), the blinking is (usually) not visible to the human eye.
stylus event is handled by the standard event handling code
in SESAME, similar to how mouse events are handled.
The software tracking the laser styli can either send all stylus events to multi-user aware applications over a socket
interface or alternatively simulate standard mouse events
for the operating system. This second alternative is somewhat limited as the operating system silently assumes a
single input device and displays only one cursor. This facility is very useful, however, as it enables the users to interact
with the standard GUI environment of the operating system
and all installed applications. Further details about the construction, the hardware, and software for MULTI will appear in a forthcoming publication.
CoViD: Merging SESAME and MULTI
Figure 5. The CoViD system in action.
The three main modifications necessary to enable the functionalities of SESAME to run on MULTI were the support
for multiple display windows, adaptations to support laser
styli as input devices, and the support for multiple simultaneous users. Together, this forms the CoViD system.
However, a limitation of the current MULTI laser styli is
that they have only a single button, while SESAME assumes a three-button input device. The middle mouse button is used for navigation, while the right mouse button is
used for extrusion, sculpting, and resizing. On the other
hand, it doesn’t make much sense to allow all users to
change the camera for the shared tabletop display as this
would allow one user to change the view while another is
manipulating content. Hence, we decide to support navigation only via a single wireless mouse attached to MULTI,
which naturally enforces a turn-taking protocol. The functionality of the right mouse button in SESAME is important
for the design process. Hence, we added a special icon to
the tool palette of SESAME, which allows users to toggle
“right-button” mode for their stylus (and also added appropriate visual feedback close to the cursor position).
First, the displays of MULTI were associated with various
views of the 3D design. The tabletop surface is best used
for a view of the design from the top, akin to e.g. the view
afforded by a map of a city. The vertical surfaces are better
suited for a “side” view of the 3D model and we typically
use two or all three wall displays together to provide a
large-scale perspective view of the design (see e.g. Figure
5). As additional information is critical in many design
meetings, we allow the users to display and interact with
additional information (e.g. a browser window or a spreadsheet) on one (or more) of the wall screens a browser window. These windows are typically placed on one of the two
“side” walls. Furthermore, several tablet PC’s are available
in the system as another means to access auxiliary information or to allow people to transition between group and individual work. Connectivity is provided via a wireless network and two USB hubs, which are mounted in convenient
locations under the table.
To support multiple operations by multiple users at the
same time, we enhanced SESAME to keep track of separate
states for each user. Also, the software needs to simulate a
separate cursor for each participant. Whenever a user selects on of the tool palette entries, his/her cursor is changed
to reflect their current mode. Furthermore, the undo functionality of SESAME was adapted by allowing people to
activate undo in any display.
Technically speaking, it was fairly easy to adapt SESAME
to support multiple views. We only had to change the rendering loop to support multiple open windows. However,
there were subtle issues related to limitations of graphics
card drivers, which prevented us from creating a single
window across that spans all three wall displays. We currently handle this case simply by opening one window on
two displays and another on the third.
Finally, we had to perform more extensive modifications to
SESAME to enable the creation of new content in parallel
by multiple users. This involved keeping track of the currently active “drawing plane” on a per user basis and modifying that information with every 2D interaction of a user.
We also adapted the 2D drawing facilities of SESAME so
that all guides are only visible in the current window – i.e.
do not show up in other windows. This avoids obscuring
other views of the 3D model with these temporary visualizations.
The second issue, supporting a laser stylus as input device,
was done by modifying the central event loop of SESAME
to check for updates from the laser stylus tracking software
of MULTI, which sends all information about stylus
movements and button presses over a socket. Then, each
7
laid on top of them. This is a compromise between the
“perfect” look of standard computer graphics and the
“imprecise” look of a pen-and-paper sketch. While it
would be desirable to provide a more “sketchy” look, the
problem is that there are no efficient and general interaction techniques that work with imprecise 3D information.
EVALUATION OF COVID AGAINST GUIDELINES
This section presents an evaluation of the CoViD system
relative to the guidelines presented in the first part of the
paper. This evaluation was performed by one of the authors,
a human-computer interaction expert. As most points focus
on one of the two main parts of CoViD (SESAME and
MULTI), we refer these parts whenever appropriate.
• Range of levels of abstraction: Snapping provides a convenient bridge between the inaccuracy of a users’ input
and the accuracy required by computer system. With
snapping, designers can quickly generate a rough scene
configuration and the system can display it accurately
(enough). In SESAME, the level of detail is proportional
to the viewing distance in many interaction techniques.
This is based on the observation that designers generally
work on an overall idea by viewing the scene from a farther distance and work on detail by zooming into the part
of interest. Furthermore, the group manipulation techniques of SESAME facilitate fluid transitions between
work at a large scale and detail modification.
• Non-intrusive interface: To minimize the cognitive load
of the user SESAME relies only a minimum set of
modes: a 2D mode for drawing contours and a 3D mode
for manipulating solids and the various tools associated
with each mode. For 2D drawing activities, the system
provides a rich set of support techniques – e.g. suggestions, automatic segmentation of freehand drawings, and
recognition of closed structures. In 3D manipulation
mode, the system provides a natural and efficient manipulation techniques based on physical properties, such
as gravity and collisions. To avoid the cognitive overhead
associated with wire-frame visualization and/or orthogonal perspective, SESAME allows the user to perform all
manipulations in a perspective view.
• Ability to edit various forms of information: Currently,
SESAME does not address this in a significant way.
• Easy creation: As solid modeling is more appropriate for
early design phases compared to other approaches such
as polygonal modeling [1], SESAME supports solid
modeling directly, via the metaphor of the extrusion of
2D contours, as well as sculpting and direct manipulation
of 3D objects.
• Supporting evaluation (simulation): We support this is
via the export of scene geometry to any simulation package (e.g. photorealistic rendering software or stress
analysis) and then running that simulation. In the future,
we may integrate this functionality better.
• Natural interpersonal interaction: As all participants are
sitting (or standing) around the MULTI table, this is similar to standard meetings around a conference table.
Hence, participants can naturally interact face-to-face
with other participants in the system.
• Easy combination and restructuring: SESAME provides
object manipulation schemes that use a single viewpoint
and match the users’ expectations about the most probable 3D object motion. Similarly, group manipulation
techniques behave in a way that is consistent with the real
world. Both of these techniques facilitate experimentation with the structure of the scene, as objects behave like
most naïve users expect them to do.
• Transitions between activities: MULTI allows its users to
interact with the standard GUI desktop, which enables
them to use many different applications and hence transition seamlessly between activities such as design, performing a WWW search, interacting with spread-sheets
or other documents, etc. As the wall surfaces are interactive as well, it is easy to transition to sessions, where one
person interacts with the system as if it were a large
blackboard. Furthermore, people can (and usually do)
place paper documents and other work artifacts on the
rim of the table, which provides even more possibilities
for transitions between various activities.
• Tolerance to ambiguity and incompleteness: Sketching
naturally supports ambiguity and incompleteness, which
makes it a great tool for the initial phases of design.
SESAME includes support for three of the most important features of sketching. First, any freehand sketch is
automatically beautified by segmenting it into primitive
2D shapes. As user input usually contains some amount
of jitter, often also caused by digitization, a certain degree of beautification is usually a positive thing. In
SESAME, the level of detail for the segmentation varies
depending on the view distance to preserve important details while removing unwanted noise. We found that this
addresses the trade-off between ease-of-use and precision
fairly well. Also, the parts of segmented stroke can be
manipulated later, i.e. if the user wants to refine a design,
which facilitates the transition to more precise designs.
Second, SESAME supports the coexistence of 2D drawings and 3D shapes, as it facilitates the “exchange” of
ideas across dimensionalities. Finally, SESAME renders
the scene as flat-shaded objects with thick outlines over-
• Transitions between personal and group work: The table
surface of MULTI is large enough that a person cannot
simply reach completely across it (without standing up
and leaning across). This effectively gives each user their
own personal space in an unambiguous manner. Furthermore, all work files are placed in subfolders of the GUI
desktop. As this desktop folder of the system is shared
via the wireless network, this makes it easy for participants to transfer (part of) the data to an external portable
device (e.g. a Tablet PC) for individual work and back to
the desktop to share it again.
8
the pool of computer science undergraduate and graduate
students (2 female, 7 male) at the local university. They
were organized in three groups of 3. Each participant was
asked to perform two tasks individually as well as collaboratively in the group. Before the experiment, we explained
how the system worked and demonstrated the use of the
laser styli as well as the relevant operations of SESAME to
each participant for about 10 minutes. At the end the experiment, users were given a questionnaire to rate preferences between working in groups vs. working individually
and which task they enjoyed the most. Statistics were analyzed via ANOVA.
• Transition between tabletop collaboration and external
work: Two USB hubs have been mounted in easy to access locations under the table. This facilitates transfer of
data from USB drives to and from MULTI. Furthermore,
the wireless network as well as the standard network
connection of the system provide for even more opportunities to transfer data to and from MULTI.
• The use of physical objects: The wide metal ledge around
the table is stable enough for a user to sit on and users
can place laptops, various printouts, notebooks, and other
personal items on it. While people may be tempted to put
a lot of stuff around the ledge, this may also block their
access to the table, which balances things nicely. The interactive tabletop itself is stable enough to support considerable weight (due to the pane of tempered glass behind it). This allows people to place objects onto the interactive part, but such objects block the projection. Currently, MULTI does not support a tangible user interface,
but we are working on adding this functionality.
Scene Assembly Task
We first evaluated the effectiveness of the system to assemble a simple object with the 3D movement technique of
SESAME. The participants were asked to assemble a chair
from parts first individually, as shown in Figure 6.
Results
When working individually, average completion time for
the task was 4 minutes and 23 seconds. When working collaboratively the average completion time was 5 minutes 40
seconds. Detailed timings are shown in Table 2.
• Accessing shared physical and digital objects: To access
a shared digital object the user just needs to point at it
with his/her laser stylus. In computer-based design applications, the orientation of artifacts on the table surface is
less of a concern, as the natural perspective for content
on the table surface is a top down view onto the content.
This makes it relatively easy to access objects from any
direction. The wall surfaces of the system provide a side
view of the 3D environment, but these surfaces have
anyways a natural “up” orientation. All physical/tangible
objects on the table surface can easily be manipulated by
reaching for them. Finally, all interactive surfaces are
back-projected in MULTI, which makes it impossible for
a user to cast shadows onto content (beyond normal
blocking of shared content with a hand).
None of the differences between the individual and group
performances are statistically significant.
(a)
• Flexible user arrangements: Due to the size and the
physical configuration of the table (only one short side is
blocked), users can sit in many configurations around it.
As the table is on heavy-duty rollers, it can even be rolled
away completely if the users only want to only use the
wall surfaces. As discussed in the previous point, in the
context of design applications, the natural view for the
table is top-down, which leaves a lot of flexibility for the
users to arrange themselves around the table.
(b)
Figure 6. Scene assembly task
(a) Initial State. (b) Target Scene.
Individual 240 191 199 440 314 480 139 231 110
Average
Individual
210
441
160
Group
300
420
288
Table 2. Scene assembly times by user (in seconds).
• Simultaneous user interactions: Due to the time multiplexing scheme, several people can simultaneously interact with MULTI. Due to the restrictions of the Windows
operating system (it only supports one cursor), standard
GUI applications do not support multiple users seamlessly. However, applications that include support for
multiple input devices can use the full functionality of
MULTI.
Design Task
In the second task we asked users to create a simple 3D
design. The task was to create 3 different 2D shapes using
the rectangle, freeform and circle tools, to turn them into
3D structures via extrusion and then create “chimneys” on
top of them. The condition was that the chimneys had to be
created using tools different from the ones that were used to
create their corresponding structures, as illustrated by Figure 7.
PILOT USER STUDY OF COVID
We conducted two pilot experiments to test the basic features of CoViD in single user mode in comparison with
collaborative mode. Nine participants were recruited from
9
Some users reported problems with the laser styli in the
palette area on the table. This seems to have been caused by
calibration problems in that region of the tabletop. The
small seam (less than 1 pixel) caused by the overlap between the images of the two table projectors was not recognized as a noteworthy issue.
The evaluation of CoViD revealed several issues that need
to be addressed, before a more comprehensive user study
can be performed. Most importantly, we need to redesign
the laser styli to be wireless and to include a second button.
Furthermore, there were calibration problems in certain
areas of the screens and some synchronization problems
between the laser diode circuit and the camera systems,
mainly due to some bugs in the camera driver. We are currently working on fixing these issues.
Figure 7. Sample configuration for the design task.
Results
When working individually, average completion time for
the task was 4 minutes and 5 seconds. When working collaboratively, average time was 4 minutes and 54 seconds.
Detailed timings are shown in Table 3.
CONCLUSIONS AND FUTURE WORK
Again, none of the differences between the individual and
group performances were statistically significant.
This paper presented CoViD, a novel system for collaborative 3D design. We first introduced a set of guidelines for
collaborative design systems. Then we introduced the new
system and presented details about its components and how
they were adapted to form a collaborative 3D design system. Subsequently, we evaluated the design of the system
against the guidelines and reported results of a pilot study,
in which users generally liked the ability to collaboratively
design in 3D.
Discussion
Due to the small sample size (only 3 groups) none of the
results of the statistical analysis were significant, which
limits our ability to draw conclusions from the data. One
factor that may have contributed to this is that the tasks
were relatively simple and the potential benefits of collaboration hence negligible. However, we can still state that
participants were able to complete the tasks reasonably
quickly.
We are currently working on addressing the issues that
were identified during the pilot study. Once that is done, we
intend to perform a more comprehensive evaluation by having a group of architectural students design building(s) for a
free lot in a collaborative setting in collaboration with a
professor in architecture. Furthermore, we also intend to
analyze how the collaborative aspect of the CoViD system
affects creativity. Another area of future work is the extension of SESAME to support additional modeling primitives
(such as rotationally symmetric and freeform objects). Finally, we are working on a tangible user interface for
MULTI and plan to extend SESAME to allow for tangible
interaction.
Individual 420 300 179 434 153 480 148 163 107
Average
Individual
300
355
139
Group
300
413
171
Table 3. Design task times by user (in seconds)
One interesting behavior we observed was that two groups
worked mainly in the perspective view and used the tabletop only when users wanted to work in parallel. This may
have been based on the relatively larger size of the scene on
the wall screens. The lack of a second button on the laser
styli was also an issue that led to unnecessary errors.
ACKNOWLEDGEMENTS
Thanks to D. Phillips, D. Dadgari, and A. Vorozcovs for
help with camera calibration and programming the laser
spot detector software, to the York Centre of Vision Research, and to NSERC for funding.
The analysis of the questionnaires revealed that on average
participants rated the four different conditions (individual
scene assembly, individual design, collaborative scene assembly, and collaborative design) about equally in terms of
how much they enjoyed the different conditions.
REFERENCES
1. J. Deisinger, R. Blach, G. Wesche, R. Breining, A. Simon,
Towards Immersive Modeling – Challenges and Recommendations: A Workshop Analyzing the Needs of Designers, Virtual Environments 2000, Springer, 145-146.
As for the collaborative assembly task, users commented
that assembling parts with others made it more confusing
because complained that other users would often interfere
and “ruin” their work. For collaborative design, users commented that working with other people made the task more
entertaining and that it was more fun to create new designs
together rather than assembling parts. This hints at the potential of CoViD to support creativity in design sessions.
10
2.
A. Druin, J. Stewart, D. Proft, B. Bederson, J. Hollan, KidPad: A Design Collaboration Between Children, Technologists, and Educators. ACM CHI 97, 463-470.
3.
M. Gross, E. Do, Drawing on the Back of Envelope: A
framework for interacting with application programs by freehand drawing, Computers & Graphics 2000, 24:835-849.
4.
M. Gross, Emergence in a recognition based drawing interface. Conf. on Visual and Spatial Reasoning 2001, 51–65.
5.
T. Igarashi, S. Matsuoka, H. Tanaka, Teddy: A sketching
interface for 3D freeform design. SIGGRAPH’99, 409–416.
6.
J-Y Oh, W. Stuerzlinger, Moving objects with 2D input devices in CAD Systems and Desktop Virtual Environments,
Graphics Interface 2005, 141-149.
7.
J.-Y. Oh, W. Stuerzlinger, J. Danahy, Comparing SESAME
and Sketching for Conceptual 3D Design, EG Sketch Based
Interfaces and Modeling 2005, 81-8.
8.
Oh, J.-Y. Desktop 3D conceptual design systems, Ph.D Thesis, 2005. York University.
9.
J.-Y. Oh, W. Stuerzlinger, D. Dadgari, Group Selection
Techniques for Efficient 3D Modeling, IEEE Symposium on
3D User Interfaces, 2006, 95-102.
12. K. Ryall, C. Forlines, C. Shen, M. Morris, Exploring the
Effects of Group Size and Table Size on Interactions with
Tabletop Shared-Display Groupware, CSCW 2004, 284–293.
13. S. Scott, K. Grant, R. Mandryk, System Guidelines for Colocated, Collaborative Work on a Tabletop Display, ECSCW
2003, 159-178.
14. J. Stewart, B. Bederson, A. Druin, A Model for Co-Present
Collaboration. ACM CHI 99, 286-293.
15. L. Shu, W. Flowers, Groupware experiences in threedimensional computer-aided design. CSCW 1992, 179–186.
16. S. Tano, T. Kodera, T. Nakashima, I. Kawano, et al., Godzilla: Seamless 2D and 3D sketch environment for reflective
and creative design work, IFIP INTERACT’03, 311-318.
17. E. Tse, S. Greenberg. Rapidly prototyping Single Display
Groupware through the SDGToolkit, Australasian User Interface Conf 2004, 101–110.
10. J.-Y. Oh, W. Stuerzlinger, J. Danahy, SESAME: Towards
Better 3D Conceptual Design Systems, ACM DIS 2006, to
appear, June 28 2006.
18. R. Zeleznik, K. Herndon, J. Hughes, SKETCH: An interface
for sketching 3D scenes. SIGGRAPH’96, 163-170.
11. A. Pavlovych, W. Stuerzlinger, Laser Pointers as Interaction
Devices for Collaborative Pervasive Computing, Advances in
Pervasive Computing, 2004, 315-320.
11
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising