We will investigate the use of constraints for the design of virtual
worlds, and in particular
for describing behaviors of autonomous agents in virtual worlds. Virtual
worlds are
becoming increasingly popular due to the definition of standard formats
for describing 3D scenes on the
web (e.g. VRML97, Java3D and the future X3D). Nevertheless, most of
these shared virtual worlds suffer
from an important drawback, namely they are not lively or interactive
enough and provide only a minimal
degree of liveliness due to the lack of autonomous creatures that
inhabit such worlds. We will therefore
consider in this paper the problem of
"populating" such worlds with
virtual agents representing life-like
creatures which could autonomously navigate and react to their changing
environment, and also possibly
interact with users. We will thus first present a general language based
on the
Timed Concurrent Constraint paradigm and also a simplified framework
with an efficient run-time
algorithm to implement those behaviors, based on local search
techniques.
As a first application of this framework, we have considered virtual
creatures with simple
reactive behaviors or limited planning capabilities inspired from
research in the field of Artificial Life
and (behavior-based) robotics.