Natural, pleasant, productive working environment (40 hours/week)
Benefit from the physical world
Benefit from the virtual world
Natural interaction, both user to user and user to objects
Users were to be unencumbered.
DesignSpace as proposed, technical manifestation
appealing, productive physical workspace, possibly a corner or U-shaped work
area tabletop.
all table and wall surfaces can display computer-generated images.
The target resolution was 25 pixels per angular degree on the retina.
For a corner desk with walls at arm's depth, this called for
2500x1500 pixels on each of two walls and 4000x1500 pixels across
the desk. This was to be achieved through tiling many displays.
the visual displays were to be bright to allow for a lighted room.
No stereo was planned because all known stereo techniques either
A) reduced the light by more than 50%, B) encumbered, or C) both.
an auxilliary immersive display device (BOOM or HMD) shall be
available to provide short-term stereo or immersive experiences.
the aural displays were to be head-tracked transaural (loud-speaker-
based binaural positional 3D audio; never before implemented.)
no computer-generated haptic display was planned, but there was
no barrier to task-specific haptic proposals. A generic haptic
display would be encumbering.
all interaction would assume human dexterity, whether or not the
specific interaction required it. Dexterous control is innately
human and without it, humans don't feel as in control. All current
functional dexterous input systems were encumbering, but several
technologies show future promise to allievate some or all
encumberance in the near future.
DesignSpace as proposed, generic functionality
the DS interface would replace the GUI of the computer OS of choice.
Assume WinNT for example, all Win32 applications would run in
DS without recompilation as a 2D paper/tablet metaphor.
All child windows could be manifested as dependent 3D objects,
including tear-off toolbars for instance. They could be manipulated
with the finger tips, a physical tool (keyboard), or a virtual tool
(mouse).
recompiled apps could become completely and inherently 3D and make
better use of dexterous manipulation.
while a DS workspace was physically a single-user environment
(unlike the responsive workbench, for instance), DS is designed as
a collaborative virtual environment. Collaborating users could be
physically at a DS station next door, or in the next country, but
virtually across the desk.
DesignSpace as proposed, design functionality
the scale of designed objects should fit within the working volume
of the DS workspace (0.8 meters cubed). An architectural piece
would be designed as 1/4" model for instance. A BOOM could be
pulled over the workspace to fly inside the architectural piece.
[ After these DS proposals, Mark Mine (of UNC at the time)
demonstrated some wonderful illustrations of changing scale in
architectural spaces. Some of these ideas could work well to
extract details, found during a BOOM fly-thru, into the DS
workspace for manipulation. ]
conceptual design can be performed dexterously, using solid model
primitives or virtual surface materials (deformable meshes).
A dexterously manipulated linking tool helps keep conceptual
assemblies together.
for all precision design exercises, a minimal, but complete set
of 3D tools (parked in consistent spatial locations about the
workspace) shall be developed for constrained operations.
mechanical design can begin with importing existing models into
the workspace to serve as constraints.
market specific design tasks such as architecture should utilize
specialized-task 3D tools ... macros of the minimal base toolset.
DesignSpace as IMPLEMENTED (most recently at CHI'94), technically
corner desk, implementing two visual display walls with pixel
accurate corner seam. An image was never made on the desktop, but
use clearly showed that a horizontal surface display was very
important. As many as three 640x480 viewplanes were tiled to
create the wall display. The images were rear projected via TFT
projectors creating pixels of 2mm square, subtending an 1/8th of a
degree of the retina of an observer at arm's length.
Polhemus Fastrak sensors were used for all 3D tracking. The head
sensor was mounted on the headset (microphone + earphones) for
systems with 3D audio, or on just a hat or headband otherwise.
dexterous input was achieved via a pair of 22-sensor CyberGloves,
with mounted Polhemus sensors. For minimal encumberance on systems
not requiring dexterous input, a button hand grip with embedded
Polhemus sensor was used.
images were generated with Division's dView 3D graphics pipeline
system, an i860-based geometry engine coupled to Toshiba raster
engines. The system was capable of 30,000 polygons/sec, but
suffered with large polygon fills. It was adequate for proof of
concept, but far too poor performing for real work. For instance,
shadow-casting is a much-needed depth cue that was necessarily
omitted.
audio, represented by voice, synthesized feedback sonification,
and sound effect playback, was rendered as emanating from its
virtual point of origin through positional binaural cue synthesis.
DS used the dynamic Head-Related Transfer Function convolution
technology of Crystal River Engineering's Beachtron processor.
The combined binaural signals of all sounds were presented over
headphones. The Beachtron's Emu Proteus wave-table synthesizer
was used to synthesize aural task feedback.
multiple systems were linked via RS-232 lines at 115.2 Kbaud,
with every system an equivalent peer. Complete frame packets were
on the order of 100 7-bit bytes depending on the activity, thus only
a kilobaud was needed for each frame per second of performance.
Plain-ordinary-telephone-system (POTS) 14.4 Kbaud modems were used
for remote communication, utilizing hardware compression to achieve
19.2 Kbaud throughput.
The internal kinematics, 3D audio, and all tracking sensors were
kept at twice the visual frame rate, but limited to less than 30 Hz.
The visual graphics framerate was limited to between 11 and 14 Hz,
depending on scene complexity.
DesignSpace as IMPLEMENTED (most recently at CHI'94), functionally
three-way remote collaboration was fully supported, including:
audio communication, object manipulation, dexterous manipulation,
aural feedback, and consistent spatial relationships. An arbitration
scheme was developed to reconcile ownership of multiply-grasped
objects, figuring grasp-quality against timestamps to overcome
communication latency.
any pre-modeled geometry could be loaded in the system as instances.
instances could be linked with a linking tool to form assemblies.
instances could be transformed (translate, rotate, scale) dexterously.
Meshes could be deformed dexterously, by translating control points.
scene graphics could be saved and reloaded as initial configurations.
pre-modeled dynamics (mass properties, elasticity, vibration, etc.)
could be loaded and applied to any instance.
DS could NOT shell any application.
DS contained NO alignment or constraint design tools.
