Motion Planning for Virtual Humans

Marcelo Kallmann
University of California, Merced

While concerted efforts in the fields of Computer Animation and Robotics have
achieved great successes in the broad domain of motion representation,
synthesis, and learning, planning complex whole-body motions among obstacles in
interactive applications remains a challenge. Once this landmark is achieved,
autonomous Virtual Humans will have the potential to impact a number of
application domains, such as training, education and delivery of instructions.
Towards this end, I propose to address the motion synthesis problem in its full
complexity, taking into account balance, coordination and planning around
obstacles. I will present in this talk an overview of my work on extending
sampling-based motion planners to humanoid characters, and will in particular
present: (1) new techniques for planning object manipulations, (2) algorithms
for planning the synchronization (concurrently and sequentially) of different
primitive motion skills such as locomotion and manipulation, and (3) first
attempts toward learning situated tasks for improving the performance of motion
planning.



Marcelo Kallmann is Assistant Professor and Founding Faculty in the Engineering
School at the University of California, Merced. He is also currently affiliated
as adjunct faculty to the Computer Science Department of the University of
Southern California (USC). Prior to joining UC Merced, Prof. Kallmann was a
researcher working on Autonomous Virtual Humans at the USC Institute for
Creative Technologies (ICT). Before that he did postdocs on humanoid robotics
at the USC Robotics Lab and on virtual human animation at the Virtual Reality
Lab of the Swiss Federal Institute of Technology (EPFL), where he completed his
Ph.D. in 2001. His main areas of interest are motion planning, computer
animation and robotics.