Poke combines a force image of a pliable planar surface with
a simple holographic video image of the same geometry. An
interacting person can both see and feel the multi-modal
 
simulation while poking the surface into an arbitrary shape.
- overview
- touch
- lathe
 
 
wendy plesniak   
 
   
  Poke has two basic subsystems: the haptics  
module and the holovideo module. The former
haptically displays a pliable surface and allows
the underlying model to be arbitrarily reshaped
by poking with the Phantom stylus. The
holovideo module propagates haptic model
changes into the hologram incrementally,
and only in regions corresponding to change.
Updated holograms are sent via HIPPI to
the holovideo module's display server
(running on the Cheops Imaging System)
which updates the holovideo display to
show the changes due to poking.
There is a 2s lag between feeling changes
in the haptic model and seeing them on the
holovideo display. And though Poke's model
can be deformed in a more arbitrary fashion
than Lathe's, its hologram update rate is
reduced to 0.3fps.
 
  Poke's force model is given by a 2D Catmull-Rom
spline whose control points are connected by a
grid of springs. As a person presses into the
surface with the Phantom stylus, the grid and
the surface near the contact point are recruited
in the resulting deformation as shown below.
In this deformation, control points are restricted
to move only in depth (z).
The holographic image shows an array of points
collocated with the haptic model control points,
and the surface is felt to stretch among them.
The holographic image is updated rapidly
using the incremental computing method which
assembles the image from a set of precomputed
elemental fringes. This method lets us make fast
and flexible local changes in the hologram.
 
A table of elemental fringes is first
computed by interfering a plane wave
with a set of spherical waves. These
spherical waves arise from a set of
points located at a finely-sampled set
of distances from the hologram plane.
Each fringe in the table reconstructs
an image of a point at the same depth
used to originally produce the fringe.
A fringe can be translated in the hologram
to translate its reconstructed image, and
a fringe can be reversed to flip the depth
of its image. Arbitrary images can be
constructed by adding appropriately
arranged fringes from the table into
a composite pattern, as shown at the
right for a simple 2-point image.
Since the hologram of any scene
modeled by a collection of image points
is simply a linear superposition of all
points' elemental fringes, we can erase
any point from the image by subtracting
its fringe from the pattem as shown.
The image of that point can be updated
by adding a new fringe to reconstruct
the point at a new location. Poke uses
this approach to update its hologram as
a person deforms the pliable surface.
This method requires operating on a
144MB "working" hologram, which is
subsequently normalized and sent to
the display server on Cheops.
 
Sponsors of this work include Honda R&D Company,
NEC, IBM, the Digital Life Consortium at the MIT
Media Laboratory, the Office of Naval Research
(Grant N0014-96-11200), Interval Research Corporation,
and Mitsubishi Electric Research Laboratories.
 
 
spi home people research publications media lab home