Generating Human Motion by Symbolic Reasoning
Files
Penn collection
Degree type
Discipline
Subject
Funder
Grant number
License
Copyright date
Distributor
Related resources
Author
Contributor
Abstract
This paper describes work on applying AI planning methods to generate human body motion for the purpose of animation. It is based on the fact that although we do not know how the body actually controls massively redundant degrees of freedom of its joints and moves in given situations, the appropriateness of specific behavior for particular conditions can be axiomatized at a gross level using commonsensical observations. Given the motion axioms (rules), the task of the planner is to find a discrete sequence of intermediate postures of the body via goal reduction reasoning based on the rules along with a procedure to discover specific collision-avoidance constraints, such that any two consecutive postures are related via primitive motions of the feet, the pelvis, the torso, the head, the hands, or other body parts. Our planner also takes account of the fact that body motions are continuous by taking advantage of execution-time feedback. Planning decisions are made in the task space where our elementary spatial intuition is preserved as far as possible, only dropping down to a joint space formulation typical in robot motion planning when absolutely necessary. We claim that our work is the first serious attempt to use an AI planning paradigm for animation of human body motion.