Individual whole body motion. Expression and generation using time-variant performance index

This paper deals with the method to generate human whole body motion by identifying performance index to be optimized. Most of the human motion in daily life, including a "lift-up an object" motion, is complex and the performance index are regarded to vary during the motion. Lift-up an object motion is generated by solving inverse kinematics. The experimentally measured trajectory of the hand is given and the rest of the degree of freedom is determined by instantaneous optimization of a performance index that blends several subfunctions. The coefficient for each sub-function is derived to fit the originally captured human motion and the tendency of the coefficients are analyzed to select appropriate sub-function(s) to estimate among several candidates.