An approach to experimentally based modeling and simulation of human motion

UnrestrictedA computational approach to estimation of the kinematic of a mechanical system consisting of interconnected rigid bodies performing motion in 3-dimensional space is developed. The main application considered in this dissertation is an experimentally-based analysis of human motion. It is assumed that an experimentally measured motion data in the form of marker locations is provided. Using the experimental data, the proposed computational procedure computes the entire kinematics of the system while preserving important physical and dynamic properties. These properties include systm integrity, and whole-body dynamic properties such as the system's center of mass kinematics, and its angular momentum.; The problem of finding the orientation of interconnected rigid bodies is formulated as a least squares problem. Constraints used to ensure the preservation of the required properties are included in the form of penalty functions. Equations for the optimal solution in the form of quaternions are derived. Since the resulting equations are highly nonlinear, embedded iteration process to compute the optimal solution was developed and implemented. A quaternion based dynamic equations for a chain of rigid body was formulated and implemented. An interpolation procedure for attitude given in terms of quaternion parameterization was developed and implemented.; Two simulation studies are performed. To evaluate the numerical efficiency of the proposed algorithm, estimation of the kinematics of a chain simulated on a computer is performed. To evaluate the performance of the proposed approach, a 14-body model is used to estimate human three-dimensional motion during jumping using experimental data. The results of the two studies indicate fast convergence of the algorithm to an optimal solution while satisfying the imposed the constraints