Human-like agents with posture planning capability

Human motion animation has applications in a wide variety of fields. For example, animating motions of the human body in given CAD models helps evaluate how well the design of the CAD models fit with human operators. Human body models are geometric structures which may be ultimately controlled by kinematically manipulating their joints, but for animation, it is desirable to control them in terms of 3D space goals of end effectors, e.g. hands. To generate human body motions to achieve such 3D space goals, we develop a motion planning method. It is motivated to handle two problems: (i) massively redundant degrees of freedom of the body and (ii) partially known goal configurations which are specified in terms of end effector goals. We reduce the degrees of freedom of the body by introducing significant "control points and vectors," e.g. pelvis forward vector, palm up vector, torso up vector such that end effector goals can be translated into their movements easily. This reduced set of parameters is used to enumerate motion dependencies among them. Motion plans for partially specified goals are suggested by incrementally constructing a goal/constraint set based on original goals, motion dependencies, collision avoidance requirements, and discovered movement failures. Global postures satisfying a given goal/constraint set are determined with the help of incremental mental simulation which uses a robust inverse kinematics algorithm. Collision-avoidance is planned using intuitive 3D space spatial relations as far as possible, rather than immediately dropping down to numerically intensive joint-space motion reasoning, whose complexity is exponential in the number of degrees of freedom.