Date of Award
Summer 2020
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Mechanical Engineering
First Advisor
Voglewede, Philip A.
Second Advisor
Harris, Gerald
Third Advisor
Moore, John
Abstract
The design and development of gait-related treatments and devices is inhibited by anabsence of predictive gait models. Understanding of human gait and what motivates walkingpatterns is still limited, despite walking being one of the most routine human activities. While asignificant body of literature exists on gait modeling and optimization criteria to achievesimulated, normal gait, particularly with neuromuscular models, few studies have aimed to applyoptimization targets which approximate metabolic cost to mechanical gait models. Even fewerhave attempted this predictively, with no joint angle data specified a priori. The Sunmodel [31], [32] is one such mechanical framework which utilizes MPC to predict the dynamics ofhuman walking. This thesis expands the Sun model [31], [32] to simulate a full gait cycle (CG) andinvestigates the application of new optimization targets within an existing Model PredictiveControl (MPC) framework for predictive gait simulation developed by Sun [31], [32] .The Sun model [31], [32] was previously limited to a half gait cycle (GC) which assumedbilateral symmetry and optimized only according to characteristic constraints such as step lengthand velocity of the center of mass (COM). In this thesis, the Sun framework and MPC controlscheme were expanded to generate consecutive double support (DS), single support (SS), DS, andSS period simulations, which constitutes a full GC. The resulting GC simulation was not markedby GC events toe off (TO) and heel strike (HS), but did achieve continuity over the period whichwas not achieved by the Sun model [31], [32] . Additionally, new cost functions were developedconsistent with existing literature which suggests that the Central Nervous System (CNS) uses avariety of energy-related targets in generating gait. This thesis demonstrates that the applicationof optimization targets which approximate metabolic costs is possible with the proposed MPCframework for a mechanical gait model, but that the performance of resulting simulations shouldnot be evaluated until a full GC marked by TO and HS is achieved.While a continuous full GC simulation was achieved, the failure of the model to reliablymeet characteristic constraints, particularly in SS, prevents simulation of a GC marked by TO andHS. The work in this thesis points primarily to the failure of the optimization routine within theMPC framework to reliably find a solution that meets constraints as the cause of this problem. Ifthe optimization problem can be classified, an appropriate solution algorithm could be chosenwhich could reliably find a solution for any given set of constraints and initial conditions (IC).Identifying an appropriate solution algorithm could make the MPC framework proposed a viablemethod of gait prediction and simulation.This investigation provides researchers better understanding of the application ofenergy-based optimization in mechanical gait models and the current limitations of gaitprediction and simulation. In addition, direction is given to the future work necessary to establishMPC as a viable control method for gait simulation.