``
Abstract:
Even though people walk everyday, locomotion is controlled subconsciously and is a very complicated motion. Getting a computer to simulate human locomotion on a graphics display is a greater challenge than might first appear. This is due to the fact that observers may not be able to characterize what makes a gait (walking pattern) "real", but they are quick to note when it is not.
This work investigates the use of rotoscoping (motion digitization) and higher-level hierarchical dynamic simulation as tools for animating human figure gait instead of traditional key-framing. Previous approaches for human figure animation are reviewed. A hierarchical high-level inverse-dynamic control strategy is discussed. Biomechanical knowledge about the locomotion cycle determines the forces and torques that derive the dynamic model of legs to produce a natural animation. Lagrangian dynamics is applied to a human figure linked-structure model to iteratively compute joint forces and torques. Only Linear path, sagittal plane locomotion is considered during the simulation. Kinematic cosmetics based on the determinants of gait are used to compensate for the coronal and transverse plans.
An implementation of rotoscoping and dynamic control is described. A variety of smooth human walks are generated with very little effort depending on few specifications, such as step length and velocity. Alot of numerical speedup measures have been employed to produce almost real-time human walking animation on PCs. For rotoscoping a marker tracker program was developed which could extract locomotion parameters and attributes from captured motion. Extracted parameters could be used to initialize the dynamic model to generate a walking sequence similar to the captured one. This research illustrates the superiority of the presented kinematic-dynamic approach, and concludes with future work recommendations.
A considerable part of this research is based on the work of A. Bruderlin and T. Calvert, "Goal-directed dynamic animation of human walking", Computer Graphics, 23(3):233-242, July 1989, and their Life Forms system.
Copies of the executables, body contours, and walking traces for the Marker Tracker program and the Walker program are available for instructionl purposes upon request.
