Knowledge-Driven, Interactive Animation of Human Running

A high-level motion control system for the animation of human-like gures is introduced which generates a wide variety of individual running styles in real time. Sequences such as a leisurely jog, a fast sprint or a bouncy run are conveniently obtained by interactively setting the values of \running" parameters such as desired velocity, step length, \ ight" height or \heel/toe" strike while observing a running gure on the screen. The algorithm incorporates knowledge of how humans run at several levels: empirical knowledge de nes the relationships between the running parameters; for example, a change in running velocity by the user triggers a change in step length to maintain a \natural" running stride. Physical knowledge calculates the trajectory of the body for the current running step. Knowledge about limb-coordination of a running stride is utilized for establishing both, state-constraints which de ne the support and ight states, and phase-constraints to \guide" the internal joint-angle interpolation for the stance and swing phases.