| Publications Page | HITL Home |
Technical Report: R-94-3
Dace A. Campbell
Human Interface Technology Laboratory
FJ-15
University of Washington
Seattle, WA 91895
and
Department of Architecture
208 Gould Hall, JO-20
University of Washington
Seattle, WA 98195
Maxwell Wells
Human Interface Technology Laboratory
FJ-15
University of Washington
Seattle, WA 91895
ABSTRACT
An addition to a building was designed using virtual reality (VR).
The project was part of a design studio for graduate students
of architecture. During the design process a detailed journal
of activities was kept. In addition, the design implemented with
VR was compared to designs implemented with more traditional methods.
Both immersive and non-immersive VR simulations were attempted.
Part of the rationale for exploring the use of VR in this manner
was to develop insight into how VR techniques can be incorporated
into the architectural design process, and to provide guidance
for the implementers of future VR systems. This paper describes
the role of VR in schematic design, through design development
to presentation and evaluation. In addition, there are some comments
on the effects of VR on detailed design. VR proved to be advantageous
in several phases of the design. However, several shortcomings
in both hardware and software became apparent. These are described,
and a number of recommendations are provided.
INTRODUCTION
The architectural design process can be broken into the following
phases: schematic design, design development, presentation and
evaluation, detail development and construction documents, bidding,
and administration of the construction. In the schematic design
phase the overall characteristics of the building are established.
Significant issues are identified, and initial design decisions
are made. During the design development phase the specific character
and intent of the entire project are described. The presentation
and evaluation phase is an iterative process during which proposals
are presented for review by a client, review board, or design
jury, and design decisions are finalized. Following the approval
of the design, details are developed and construction documents
are produced. These may be a combination of working drawings and
written specifications which serve as a legal description of what
is to be built. As the construction documents near completion,
they are released for bidding, and a contractor is selected. The
final phase of the design process is the one in which the architect
administers the construction, interpreting changes and judging
perfomance [7].
Throughout all of these phases, architects find themselves perfoming
a variety of tasks, ranging from the most creative to the utterly
mundane. Computers were introduced to the architectural profession
with the hope that they would free architects of the mundane,
manual tasks, as well as aid in the management of information.
Use of Computer-Aided Design (CAD) has grown over the decades.
It has aided in the automation of tasks and in the management
of information, especially in the later phases of the design process.
However, CAD has had little impact on the earlier phases of design.
Thus, there is a point in the design process when architects and
designers must make a mental leap from sketches and study models
to CAD representations in two or three dimensions.
Efforts are being made to encourage the development of CAD systems
to enable their use by architects earlier in the design process
[5,6]. An important prerequisite for the increased acceptance
and use of CAD is an interface which will allow architects to
create and interact with their digital designs more intuitively.
Virtual reality (VR), perhaps the most advanced of three-dimensional
interfaces, has much potential for enhancing the way architects
and designers interact with their digital models [2,4].
VR has been proposed as a useful new tool for architects and designers
[3,8]. It is recognized that most of these benefits (and subsequent
use of VR by the design professions) will occur only after further
advancements of the technology [1]. However, the specific advancements
that are required can only be identified and implemented after
extensive use of the technology. This iterative cycle of use,
assessment, redesign, and use results in tools which are better
suited to the job. The method of redesigning tools by observing
how they are used is a common one among ergonomists and human
factors professionals. The rationale is that accomplished users
are best able to recommend and assess changes.
The goals of the project described below were to explore how architects
can use today's virtual reality technology in the early stages
of the design process, and to gain insight into its advantages
and shortcomings. From this insight, it was envisioned that specific
recommendations could be made for advancements of the technology.
In addition to a general observation of the use of VR in the design
process, our exploration focused on four issues. These were:
The effect of the type of interface (immersive or non-immersive)
on the designer's ability to study the design.
The effect of the level of abstraction of a complex 3-D space
on the perception of that space.
The utility of VR as a design medium during the earlier phases
of the design process.
The utility and acceptability of fly-throughs as a tool for representing
and presenting architectural designs.
METHODS
To explore these issues, we analyzed a design project implemented
by a graduate student of architecture. The eight-week project
was carried forward using available VR technology. The design
project was an addition of a conference room and exhibition gallery
to a building on the campus of the University of Washington and
was part of a design studio for graduate students in architecture.
Typical of most student projects, the designs from the class were
developed from schematic design, through design development, to
presentation and evaluation. It was not intended that the project
cover into detailed development or the production of construction
documents. The design represented with VR technology was compared
to designs generated by other students, whose projects were designed
by hand and with traditional 2-D and 3-D CAD.
PROCEDURES AND APPARATUS
In its earliest stages, the design was developed with sketches
and small physical models. The information from these was input
into a CAD modeling program. The database from the CAD program
was then exported and converted for use in real-time fly-throughs
with the VR technology. These simulations were recorded onto VHS
tape for record keeping and for further study of the design. A
detailed journal of the design and simulation processes was kept
for later analysis.
The CAD software used was formZ (from auto.des.sys), a three-dimensional
modeling software package hosted on a Macintosh computer. Data
Interchange Format (DXF) files were exported from formZ and converted
into two different formats on a weekly basis.
In one process, the DXF files were converted to Description of
Geometry (DOG) files. These files were simulated in real-time
with software in development at the Human Interface Technology
Laboratory. The simulation software was run on a DEC Alpha 600
with a Kubota Denali 6/20 graphics board. The simulation was viewed
on a high-resolution monitor (19", 1280x1024 resolution).
Navigation through the virtual environment was accomplished with
a Spatial Systems Spaceball, which allowed control of motion in
six axes. These fly-throughs were not immersive.
In the other process, the DXF files were loaded into Autodesk
3dstudio (3DS), hosted on a DOS platform, and saved as 3DS files.
These files were then converted to script (MAZ) and geometry (VIZ)
files for use by dVISE software from Division. These converted
files were then used in immersive simulations on a PROvision 200
system, also made by Division. This simulation was viewed on their
Virtual Research Flight Helmet (360x240, 90-degree field of view).
Head motion was tracked magnetically using the Polhemus 3Space
Tracker. Movement of the head caused appropriate movements of
the visual information on the Head-Mounted Display (HMD). Navigation
through the virtual environment was accomplished by pointing the
head in the intended direction of travel, and pressing a button
on a hand-held wand.
THE DESIGN PROCESS
Schematic Design
A proof of concept demonstration was first attempted while the
design was in early schematic development. During this earliest
phase of an architectural design there are only rudimentary ideas
to be represented. Therefore a "massing" model of the
design, one in which only the basic forms of a design are represented
without detail, was generated, translated, simulated and recorded.
Once the DOG files were created and some initial changes were
made to the model's orientation in the coordinate system, we added
three lights to the simulation: a blue ambient light representing
the sky, a yellow directional light representing the sun, and
an orange point-light source tracked to the participant to aid
in the perception of distance. A wireframe grid was also added
to give a sense of horizon, to aid orientation in this initial
environment. At this stage, the simulation was viewed with an
untracked Optics-1 HMD, but the small field of view (23 degrees)
and lack of head tracking did not allow adequate assessment of
the design.
Design Development
The design was developed over the next several weeks, and each
week a fly-through was conducted and recorded. It became apparent
that the delay between the conception and visualization of design
ideas did not provide direct or immediate feedback in the design
process. However, the simulations did provide a way to examine
the CAD model, to detect flaws in its construction. The simulations
allowed the opportunity to evaluate design elements such as proportion,
scale, and order; these things were not immediately apparent to
designers using CAD models alone.
As the model was developed, the frame rate of the simulation dropped
from fifteen Hertz at the beginning to about five Hertz. A significant
challenge in the design process became the issue of level of detail.
In order for the simulations to be of significance to the designer,
the model had to be developed in such abstraction that the frame
rate of the simulation was reduced appreciably. Decisions about
how to abstract the design were made by the designer, based on
aesthetic judgement and design sensibility.
About mid-way through the process, we enhanced the realism of
the representation. The horizon grid was replaced by massing models
representing the urban context of the design. The CAD database
was organized into layers, and by exporting the CAD model by layers
as multiple DXF files, we were able to assign unique colors and
transparencies to an otherwise opaque and monochromatic model.
These colors and levels of transparency were adjusted in real
time using a dial box to adjust color (RGB) and transparency (alpha)
values. Additive transparency was used to represent glass objects
because other transparency algorithms (such as subtractive) proved
to slow the frame rate to unaccepable levels. Texture mapping
was also attempted, but the textures vibrated (swam) in the simulation,
due to lack of resolution in the system (floating-point round-off
error). Interactive section cuts, which would provide the capability
to cut sections through the model with clipping planes in real
time, were also considered. However, there were problems with
getting the clipping planes to operate as desired.
Abstract elements of trees, furniture, and people (entourage)
were added to the model to enhance the sense of scale. After experimenting
with flat-shaded and wireframe representations, we found that
by making the furniture transparent, the degree of which could
be controlled in real time, it enabled us to evaluate the spatial
implications of the design with and without the furniture. This
was a useful design feature. Finally we added live video footage
as texture maps, to represent the location and character of display
and projection screens in the design. In this case, the swimming
of these textures was imperceptible due to the motion of the video
footage itself.
Detailed design
The use of VR early in the design process forced the detailed
development of the interior space as much as the exterior. By
having the opportunity to "go inside" the design and
see it from within, the designer was forced to solve complex connections
and details which would not have been apparent with other media.
The design developed much more than those of other students not
using VR as a design medium. With VR, the designer had to develop
the entire three-dimensional model to a convincing level of detail,
whereas other students concerned themselves with only specific
views and details.
Once the model was colored and detailed such that there were more
than 10,000 polygons to be rendered, the simulation slowed to
unacceptable frame rates (3-4 Hz). In order to continue to develop
the design in greater detail, a separate model was generated representing
a portion of the design. This second model was then developed
to a high level of detail not easily accomplished by traditional
architectural modeling methods. When this was simulated, we found
that the Spaceball and monitor (non-immersive VR) aided in the
perception of details and connections, but it was quite difficult
to maneuver in tight spaces. It was necessary to view the model
more intuitively so that the details and connections could be
more easily studied. At this point, we attempted immersive simulation
with a tracked HMD and wand. This was a whole new paradigm for
evaluating spatial qualities of the design. The frame rate was
extremely low (1-2 Hz) and therefore quite disorienting, but we
were able to inspect details and connections quite competently
by having more intuitive control over the viewpoint.
In both the immersive and non-immersive VR, flying through the
design, as opposed to walking through it, had some advantages
as well as some disadvantages. Flying provided a means of adopting
viewpoints that could not be easily achieved in the real environment.
This was useful for inspecting interior details, or for evaluating
the exterior of the building from a number of viewpoints. However,
there was a certain loss in the sense of scale due to the absence
of any effort required to move locations. This suggested a need
for some type of treadmill to improve the navigational interface.
Presentation and Evaluation
This project included not only a study of architectural representation,
but also of presentation. The VR system involved was not available
for the presentation and evaluation of the design, so the real-time
simulations were recorded weekly onto VHS tape for periodic review
by design critics. The video was presented to the instructor of
the design studio on a weekly basis, as well as to guest design
critics (juries) throughout the duration of the project.
Initially, we found that the critics were unable to successfully
critique the design with the VHS tape alone, because it was displaying
a walk-through without pausing on certain aspects of the design
which merited discussion. In later walk-throughs, we paused at
specific views and details, anticipating that the critics would
prefer to discuss those particular aspects. These presentations
were also supplemented with drawings and still frames of the simulation,
to allow the critics to refer to them as the video of the walk-through
moved on. Unlike other student presentations, no physical presentation
model was built. When the presentation consisted of a mix of video
footage and still images, the critics were then able to succesfully
critique the design.
DISCUSSION
VR is already a useful tool in the design process. There are,
however, issues which need to be addressed as VR technology is
integrated into the design fields:
Immersive vs. Non-Immersive VR
Both immersive and non-immersive VR were useful in the design
process. Immersive VR, with a tracked HMD and wand, offered the
designer a better perception of space and the opportunity to see
the design from the inside. At the scale of a person within the
building, the designer was able to examine details and connections
more intuitively with an easy-to-control viewpoint. This became
very useful later in the design process, as the designer was able
to detect minor flaws in the model.
Non-immersive VR, with a monitor and spaceball, offered higher
resolution and higher frame rates, both of which became necessary
as the model increased in complexity. The non-immersion offered
easier and quicker manipulation of the viewpoint. This was useful
for moving around the exterior of the building for fly-throughs
for presentations.
Level of Detail
Once a critical threshold of detail was represented in VR, the
designer was able to perceive spatial characteristics of the design
that may not have been apparent with other design media. Before
the complexity of the model reached a certain level, the use of
VR as a design tool seemed to be a viable, but not a unique, tool
of representation.
The real-time simulations became more useful as a design tool
as the level of detail of the model (color, transparency, and
geometric complexity) increased. However, the level of detail
needed to be kept in check to keep the frame rate at an acceptable
level. The challenge presented by this conflict required both
the generation of a second, more detailed model, and the skills
of the designer to abstract the models. Although more powerful
geometry engines are continually being developed, it is unlikely
that we will ever be satisfied with the level of detail that can
be simulated in real time. This may indicate a need for new ways
to display complex geometry to the viewer, both in terms of rendering
algorithms and in terms of the arrangement of the database.
New algorithms need to be developed such that complicated objects
can be displayed with a certain level of abstraction. Currently,
designers and modelers must make decisions about which aspects
of a design are most critical to its character, and remove polygons
which are not. This process is time consuming and subjective,
relying on skills and intuitions of the designer or modeler. It
would be more beneficial to design in digital media if there were
algorithms to simplify geometry yet maintain its aesthetic character.
In addition to such abstraction in the display of complex geometry,
the database could be arranged in a "hyper-geometry"
format. Such a format could be developed so that higher levels
of detail are represented only when a specific portion of a design
is being studied. With the database arranged in such a manner,
a designer in a room, for example, could pick a specific object
or detail, and be presented with more information (geometric and
alphanumeric) about that particular condition of the design. Such
a "hyper-geometry" format would be consistent with the
way architects are accustomed to representing their designs, with
overall views as well as blow-ups and studies of typical and atypical
details.
Immediacy of the Medium
Because the simulations were perfomed weekly, after exporting
and translating files from the CAD database, there was no immediate
feedback from the walk-throughs. Although VR was useful for evaluating
the three-dimensional model, it was not useful during the conceptualization
of the design. Instead, the design was developed using a combination
of sketching and three-dimensional CAD modeling, followed by simulations
in real-time.
If an inclusive modeling package was available to designers during
conceptualization of design ideas, digital models could be generated
in much the same way that physical models are constructed to enhance
the perception of a design developed by drawing. If the VR medium
could provide the immediate feedback of CAD or more traditional
design media, it is entirely possible that VR could replace "modeling"
as CAD is replacing (or has replaced) "drafting."
Two-dimensional media offer a limited means of representing three-dimensional
space. Three-dimensional media enhance the perception of three-dimensional
space. Designers need a digital design medium which allows them
immediate, direct, and more intuitive control over their three-dimensional
design, and VR can help. An inclusive, three-dimensional, world-building
toolkit that matches the sophistication of today's CAD software
would supplement, but not replace, other design media. Such software
is sorely needed before VR can significantly enhance the design
process.
Presentation Tool
The videos of the walk-throughs to present this project gave design
critics the opportunity to visualize the design as it developed.
They replaced the need for physical models and made clear what
was not apparent in CAD drawings. The response from design juries
was very positive, much more so than expected. Design critics
often found the taped "walk-throughs" very convincing
and conveyed that the design would make a very believable building.
This says much about VR as a presentation tool; professionals
and design critics, not just clients and laypersons, are able
to visualize ones design intentions more clearly with VR than
with traditional means of representation.
However, several design critics and jury members commented that
they would have gotten more out of the experience had they been
able to walk or fly through the design themselves rather that
depend on views from a particular path flown for the presentation.
A VR system was unavailble to them because of the high cost and
complexity of moving it to the place of presentation. Even if
the system was available to them, it would have been time consuming
for each of them to walk through the design individually, and
awkward to discuss the design with others not experiencing the
simulation in three-dimensions. Clearly, this problem could be
addressed by the introduction of an inexpensive, multiple-participant
VR system.
RECOMMENDATIONS
Develop an HMD with higher resolution.
Develop methods for moving through virtual environments which
do not contribute to a decrement in the sense of scale of space.
Develop algorithms which can abstract detailed geometry in a way
which retains the aesthetic character of the represented object.
Develop a "hyper geometry" format for management and
display of complex databases.
Develop sophisticated, immersive world-building toolkits for use
by architects and designers in virtual environments.
Develop inexpensive VR systems that accomodate multiple participants
and real-time communication.
ACKNOWLEDGEMENTS
We would like to thank Marc Cygnus for his help with programming
and the real-time simulations. We would also like to thank instructors
David Miller, Brian Johnson, and Jim Davidson of the College of
Architecture and Urban Planning at the University of Washington
for their advice and insight in this project.
REFERENCES
1 Airey, J., Rohlf, J., & Brooks, F. P. Towards Image Realism
with Interactive Update Rates in Complex Virtual Building Environments.
(25-28 March 1990). Computer Graphics: Proceedings of 1990 Symposium
on Interactive 3D Graphics, 24(2), pp.41-50.
2 Astheimer, P., Felger, W., & Muller, S. Virtual Design:
A Generic VR System for Industrial Applications. (November/December
1993). Computers & Graphics, 17(6), pp. 671-677.
3 Brooks, F. P. UNC Walkthrough. (August 1989). ACM SIGGRAPH `89
Course Notes: Implementing and Interacting with Real Time Microworlds,
Course 29.
4 Brooks, F. P. (1993). Virtual Reality--Hype or Hope: What's
Real? In Proceedings of the IEEE Symposium on Research Frontiers
in Virtual Reality (p. 3). Los Alamitos, CA: IEEE Society Press
5 Furness, T. A. (1987). Designing in Virtual Space. In W. B.
Rouse, and K. R. Boff (Eds.), System Design. Amsterdam: North-Holland.
6 Lansdown, J. (1994). Visualizing Design Ideas. In L. MacDonald
and J. Vince (Eds.), Interacting with Virtual Environments (pp.
61-77). Chichester, England: Wiley.
7 McGinty, T. (1979). Design and the Design Process. In J. C.
Snyder and A. J. Catanese (Eds.), Introduction to Architecture
(pp. 152-190). New York: McGraw-Hill.
8 Schmitt, G. (1993). Virtual Reality in Architecture. In N. M.
Thalmann and D. Thalmann (Eds.), Virtual Worlds and Multimedia
(pp. 85-97). Chichester, England: Wiley.