Date: Mon, 17 Feb 1997 18:10:56 -0800 (PST)
From: CS Anonymous FTP <ftp@kohler.CS.Berkeley.EDU>
Message-Id: <199702180210.SAA06704@kohler.CS.Berkeley.EDU>
To: ddgarcia@cs.Berkeley.EDU
Subject: WWW form from brunello.cs.berkeley.edu

VRP01_NAME1=David Parks
VRP02_NAME2=N/A (this is part of a research effort)
VRP03_TITLE=An Architecture for Design and Manipulation of Autonomous and Directable Agents
VRP04_DESCR=An autonomous agent is one that is capable of intelligent
and believable behavior without the intervention of a
designer.  A directable agent is one that is capable of 
manipulation at many levels of detail but is capable
of believable behavior without it.  A number of attempts
have been made at designing an efficient structure for
the creation of both types of characters.  To date, however,
few have viewed the two types and being related.

I will construct interfaces for the description of
complex animated structures.  In particular, I will :

*) Provide the user with an effective method for describing
a character as a system of limbs and joints.  Note, for
example, that when a human being is modelled it is often
done as a system of about 20-30 limbs connected thru joints
with simple range resrictions.

*) Implement inverse kinematics and other control
routines (spline interpolation and other animation
standards) for that the user will interface with
an animation system.  In other words, the user will
be able to describe complex actions thru a graphical
interface that is connected to various controllers.

*) A simple library of percepts will be available to the
designer, possibly with the ability to construct more.
The agent will need the ability to determine what it
can see/can't see, what it can hear/can't hear, etc.

*) An action selection mechanism will be internally
constructed.  The user will have access to it at a high
level; the designer will be able to effect decisions
in ways that will appear similar to emotions.  Action
selection architectures for this purpose are already in
place but I will likely use a more basic system
as the benefit of others has not been shown.

*) The structure for an autonomous character will then be
in place.  I will extend the mechanism with a video-
editing interface so that the user may dynamically
modify the actions of the agents in the classical animation
style.  As an example, I will create an animation of a
battle-scene that takes place with skeletons as actors.
The basic actions are likely to be simple; some permutation
of walk, run, swing_sword_@_x, block_x_with_shield, etc
will likely suffice.

Time permitting, I would like to export a library of
routines for use in a program so that the user might
be able to simply place the creation into his/her
own world.  The logistics of this have not yet been
considered.