From: harrison@beta.lanl.gov (David A. Harrison)
Subject: SCI: Reducing # of objects to sizeable chunks
Date: Fri, 31 Jul 1992 15:46:20 GMT
Message-ID: <1992Jul31.154620.2170@newshost.lanl.gov>
Organization: Los Alamos National Laboratory



Dear all VR enthusiasts,

Someone asked a question about how to divide the virtual world into
sizable chunks in order to reduce processor time.  I've thought about
this for a little while.  It seems that most ways have been discussed
in one manor or another.  What I propose here is in addition to any
other approachs not a replacement for them.

Besides keeping track of containment (i.e. I completely contain my
innerds therefore there is no reason to make the outside world aware
of them) for individual objects, the designer can use the containment
to consciously divide the world to reduce object interaction.

One approach is to make partial containers seem like complete
containers.  For example a room might have a window.  As soon as a
room has a window, you have to deal with rendering the outside world.
Fortuntately the outside world won't usually change very quickly.  You
could render it once (or once per long period of time) and then place
the single rendering on the wall like a painting.  When the window is
not rendering the outside world you make it opaque as far as object
interaction is concerned.  Once the window becomes opaque you don't
have to deal with rendering the thousands of objects that you might be
able to view from the window.

Compressing objects into planar paintings will work elsewhere to
lessen the load on the computer.  If you are viewing stationary
objects from a long distance, you can force the stationary objects to
become a single renderable object--a flat painting.  With long
distances we lose perspective anyways.  Furthermore you can update
this painting of objects that are far away only once every so many
seconds or so much change in angle of viewing.

Here are three other approachs that I've seen posted that had to do
with reducing the amount of overhead.  Lose detail over distance by
not using high resolution coordinates or by using some sort of
averaging.  Secondly, use less detailed versions of the same object as
it gets further away.  Thirdly, use one of the above two approachs to
reduce detail in objects that are either not in use or are in your
peripheral vision.
 
Sincerely,
David Harrison