Richard Thurman (940422.1600)--
By the way -- how would you feel about putting your arm
pointing demo into a VR? It might be an interesting demo at
some VR conferences. The VR community is just now getting
around to thinking about how to create virtual citizens.
Feel free to use it in any way you like. Version 1 is probably all
you need: source code available on request (I think it's in Turbo
Pascal).
Great! I will begin to prep our folks for "a little extra work."
Basically I need to get one of our programmers to free up some time
to kinda do this on the side. He will probably have to convert it
all to "C" and may show some reluctance to do it. It will have to be
one of those slow "I'll do it when I get to it" kind of projects.
Think what it will look like though. Imagine of donning a VR
helmet and walking into a virtual room. You see seated before you a
stick figure of an individual. You place your finger upon a plane
of virtual glass and immediately the figure places its index finger
opposite yours. Move your finger around... the figure tries to match
your movements. A fun little demo, no?
I've thought that one way to make gobs of money for PCT would be to
convince the people at Industrial Light and Magic that the way to
create computer monsters and people is to simulate their behavior
instead of animating it. Even Little Man Version 1 behaves more
realistically than animations do. If someone would supply me with
the forward dynamical equations for an animal with a body and four
two-jointed legs, I would like to write the control systems for
walking, creeping, jumping, etc.. I'll bet that the result would be
very much like a real organism.
I'm afraid that supplying you with "forward dynamical equations" is
a little out of my league. Do such equations exist? If you can
give me the correct information (buzz words, people to call, sources,
etc. ) I may be able to help obtain them -- but I gotta know how to
find them.
The human eye is awfully good at
seeing the flaws in animations, even when they're traced from life.
But how do I convince IL&M that they should hire me a physicist-
mathematician for six months to develop the dynamical equations?
Would anybody in your outfit be interested?
Obviously, I would be very interested. However, I don't got no
money.
Behind the scenes here I have been trying figure out some ways to
create a PCT research agenda. For example, last August I got a ten
minute notice to submit an ILIR (Independent Lab, Independent
Research) proposal. ILIR's are designed to allow a researcher do
'something near and dear to his/her heart' that lies outside the
scope of their current research assignments, but will add to the
research base. Sometimes some money 'magically' becomes available
to do such things. Its very volatile and one cannot count on getting
it. On this occasion I had minutes to write a one page proposal to
get some of these funds. All I had time to do was grab the "Welcome
to CSG-L" document and try to cut and paste a proposal together. I
submitted the following:
ยทยทยท
----------------------------------------------------------------------
----------------------------------------------------------------------
Title: Perceptual Control of Motion in Virtual Environments
Scientist: Dr. Richard Thurman (AL/HRAU)
Cost: Total = $150K
Hardware= $120K
Labor= $30K
Innovation:
While the existence of control mechanisms and processes (such as
feedback) in living systems is generally recognized, the
implications of control organization go far beyond what is generally
accepted in the behavioral sciences. For example, most behavioral
research is based upon a stimulus-response model of behavior. In
addition, most of the cognitive sciences approach behavior from a
command-driven model. However, in at least one instantiation of
control theory, Perceptual Control Theory (PCT), behavior is based
neither on stimuli nor on commands. Instead, behavior is posited to
be the control of perception.
PCT requires a major shift in thinking from the traditional approach:
that what is controlled is not behavior, but perception. Modeling
behavior as a dependent variable, as a response to stimuli, provides
no explanation for the phenomenon of achieving consistent ends through
varying means, and requires an extensive use of statistics to achieve
modest (to the point of meaningless) correlations. Attempts to
model behavior as planned and computed output can be demonstrated to
require levels of precise calculation that are unobtainable in a
physical system, and impossible in a real environment that is changing
from one moment to the next. The PCT model views behavior as the means
by which a perceived state of affairs is brought to and maintained at a
reference state. This approach provides a physically plausible
explanation for the consistency of outcomes and the variability of
means.
We propose to explore the use of PCT in creating more robust models of
human performance in virtual environments. In particular we propose to
construct synthetic humans (existing only in a virtual environment)
which possess characteristics of human behavior which have eluded
modelers in the past. PCT based models of human behavior will be
constructed and compared with conventional models (such as the University
of Pennsylvania+s JACK system which is based upon inverse kinematics.)
The basis of comparison will be how well these synthetic models
correspond to humans performing the same tasks in a controlled lab
environment.
Discussion:
This research is decidedly different. If successful it will provide a
jumping off point for a major paradigm shift in the behavioral sciences.
It will show that there is a plausible third alternative to explaining
behavior in human functioning and will provide computer modelers with a
new way to create computer generated actors which behave in realistic
ways.
----------------------------------------------------------------------
----------------------------------------------------------------------
The ILIR funds dried up before the proposal ever got evaluated. Too
bad.....
Rich
--------------------------------------------------
Richard Thurman
Air Force Armstrong Lab
6001 S. Power Rd. BLDG. 558
Mesa AZ. 85206-0904
(602) 988-6561
Thurman@hrlban1.aircrew.asu.edu
---------------------------------------------------