The main objective of this bachelor thesis in Industrial Technology Engineering is to predict dynamically
consistent walking motions from kinematic and dynamic measurements obtained at the
UPC Biomechanics Laboratory. A healthy gait cycle is captured and foot-ground contact forces are
measured. Then, in order to acquire the new motions, optimal control techniques are applied.
The human body is modeled with a multibody system formed by rigid bodies. Concretely, a twodimensional
simpli ed skeletal model focused on the lower extremity is used in this work. It is
formed by a total of 12 rigid bodies (trunk, pelvis and leg segments) and it has 10 degrees of freedom.
The inverse dynamic analysis is performed using OpenSim, a free software tool developed by
Stanford University that is widely used by the scienti c community.
The optimal control algorithm to obtain dynamically consistent walking motions from experimental
data is implemented in MATLAB. Moreover, the software used to solve the optimal control problem
is GPOPS-II, a general-purpose MATLAB-based software for solving multiple-phase optimal
control problems, developed by the University of Florida. Parameters of GPOPS-II are changed to
study the in
uence on the solution. Then, di erent formulations are analyzed to assess convergence
and similarity between the new motion and the captured one.
During this report, all the processes involved in the analysis and the related theory are detailed,
as well as the methodology used. Theoretical background is presented and complemented with
examples of other works. The skeletal model used is described in detail. The process to export
and obtain the experimental kinematics and dynamics using OpenSim is explained step by step.
Optimal control theory and GPOPS-II working environment, which are employed as the tool to
predict new motions, are also explained. And nally, results are presented and discussed.
This project is considered an initial study of optimal control techniques to predict human motion.
Thereby, it allows to understand these techniques and gain knowledge about how they can be used
in order to be applied, in the future, in more complex models.
