             
|
|
|
Computer Graphics Forum
Volume22, Issue 3 (September 2003)
Authors:
Marcelo Kallmann
1 Robotics Research Lab, University of Southern California, Los Angeles, United States,
kallmann @usc.edu
Amaury Aubel
2 DreamWorks Animation, Glendale, United States,
aaubel@anim.dreamworks.com
Tolga Abaci
3 Virtual Reality Lab, Swiss Federal Institute of Technology, Lausanne, Switzerland, (tolga.abacildaniel.thalmann) @epfl.ch
Daniel Thalmann
3 Virtual Reality Lab, Swiss Federal Institute of Technology, Lausanne, Switzerland, (tolga.abacildaniel.thalmann) @epfl.ch
Summary:
We present new techniques that use motion planning algorithms based on probabilistic roadmaps to control 22 degrees of freedom (DOFs) of human-like characters in interactive applications. Our main purpose is the automatic synthesis of collision-free reaching motions for both arms, with automatic column control and leg flexion. Generated motions are collision-free, in equilibrium, and respect articulation range limits. In order to deal with the high (22) dimension of our configuration space, we bias the random distribution of configurations to favor postures most useful for reaching and grasping. In addition, extensions are presented in order to interactively generate object manipulation sequences: a probabilistic inverse kinematics solver for proposing goal postures matching pre-designed grasps; dynamic update of roadmaps when obstacles change position; online planning of object location transfer; and an automatic stepping control to enlarge the character's reachable space. This is, to our knowledge, the first time probabilistic planning techniques are used to automatically generate collision-free reaching motions involving the entire body of a human-like character at interactive frame rates.
Categories and Subject Descriptors (according to ACM CCS): I.3.7 [Computer Graphics]: Three-Dimensional Graphics and Realism