Improving Behavioral Animation Through Machine Learning
Jonathan Dinerstein
Provo, Utah
Parris Egbert
Provo, Utah
Project Description:
In the field of computer animation, there is
widespread interest in creating intelligent virtual characters. An intelligent
character is an autonomous agent and can animate itself, alleviating a human
animator of a large workload. This is especially pertinent in interactive
virtual environments (such as training simulators and computer games), where
the behavior of a character cannot be dictated in advance. In our research, we
are developing learning techniques for virtual characters, such that they
automatically learn how to perform their assigned tasks and adapt to more
optimally interact with human users.
Below are links to our work. Both PDF versions of the papers and digital videos
are available. Other research we have
performed (which is not related to the topic of behavioral animation) can be
found here.
Published research:
·
“Fast and Learnable Behavioral and Cognitive
Modeling for Virtual Character Animation”.
To appear in Journal of Visualization and Computer Animation, 2004.
Submitted research:
·
“Fast Multi-Level Adaptation for Interactive
Autonomous Characters.” Submitted
to SIGGRAPH 2004 (ACM Transactions on Graphics).
·
“Fast and Robust Incremental Action Prediction
for Interactive Agents.” Submitted
to Computational Intelligence.
·
“Intelligence Capture: Automatic Behavioral
Animation from Human Example.” Submitted
to Journal of Graphics Tools.
Example Pictures:
These figures are composed of sequential frames from animation sequences generated using the algorithms we have developed.
