A fundamental question in biomechanics is why we walk the way that we do. Given a wide range of options to transport the body overground that include hopping, skipping, and running, at slow to moderate speeds most people choose to walk, and they do so in stereotypical ways. One possible explanation for the consistency of movement patterns is that the central nervous system selects movements that optimize one or more performance criteria that represent high-level goals of the movement. Commonly considered performance criteria are the metabolic energy required to complete the task, maintaining adequate stability, and the vigor with which movements are performed. The relative priority placed on these criteria in generating movements is difficult to directly control in an experimental setting, but can be readily modulated using musculoskeletal computer modeling and simulation. While some of these criteria have been studied extensively (e.g., metabolic cost) others, such as movement vigor, have only recently gained widespread attention in the literature. It seems likely that multiple performance criteria are important in gait selection. However, this presents a challenge to our analysis and understanding as the relationships among criteria grow combinatorially with the number of criteria considered, making the problem computationally intractable. Understanding the influence of performance criteria on gait is further complicated when we consider physiological changes to the musculoskeletal system due to aging or disease, such as muscle weakness.

This dissertation consists of three related studies targeting various aspects of performance criteria and their effects on human gait. First, we explore the relationship between movement vigor and metabolic cost in human gait speed selection. Our findings provide a rationale for human gait speed selection as a trade-off between the contrasting tendencies to minimize energy expenditure and to discount long task completion times. In the second study, we analyze the relationships among a broader range of performance criteria with both quantitative and qualitative aspects of human gait. This broad characterization allows us to develop a novel mapping between the importance placed on performance criteria and gait behavior. In our final study, we decouple the effects of performance criteria from changes to the musculoskeletal system representing bilateral and unilateral muscle weakness to examine how each contributes to gait selection. We employ a combination of musculoskeletal simulation, Monte Carlo methods, and machine learning techniques to address the complexity of the multi-criteria problem to better understand gait behavior.

Performance criteria, such as minimizing metabolic cost, have long been thought to influence our movement patterns. This dissertation provides unique insights into how performance criteria affect human gait using a predictive musculoskeletal simulation approach. In particular, we demonstrate for the first time the ability to establish mappings between the relative priority placed on a variety of performance criteria and both quantitative and qualitative characteristics of human gait. This approach has the potential to:​ 1) inform our fundamental understanding of human gait and 2) provide a basis for improving human performance. For example, developers of robotic assistive devices are seeking to embed musculoskeletal models into the control algorithms for these devices. To realize this goal, it will be necessary to have a well-established mapping between the high-level goals for the user and the movement patterns that satisfy those goals. This work represents an important first step in achieving this objective.