Kinematic and dynamic techniques for analyzing, predicting, and animating human locomotion

In this work, a visually realistic and dynamically sound animation of human locomotion is obtained using both kinematic and dynamic techniques. Even though the macro-physical world can be predicted quite accurately by the physics law, we manifest that dynamic techniques alone can not predict the behavior of a self actuated system. We present a technique in which kinematics and dynamics are coupled together to form the cerebrum-cerebellum animation mechanism. Kinematic techniques are used to initiate and generate goal oriented intentional motion, and dynamic techniques are applied at each frame to adjust the kinematic motion to achieve dynamic soundness without affecting the goal achievement. In the locomotion case, dynamic soundness is judged by dynamic balance of the overall system and motion comfort which is a comparative measure of the current joint torques with human strength. The dynamic control module DYNCONTROL performs balance and comfort control based on inverse dynamics and strength data. For the inverse dynamics computation, the Newton-Euler method is applied to a 97 degree of freedom human body model to compute the joint forces and torques in real-time. The effect of attached load or external forces is simulated quite accurately. Kinematic locomotion generation module KLOG is responsible for various locomotion primitives and rhythmic and non-rhythmic variation of such primitives, and cover curved path walking and running, lateral, backward, and turnaround steps, the transitions between walking and running for motion continuity, etc. Our technique has been applied to the areas such as reactive incremental path planning and virtual reality.