Walking is the primary way in which humans move around in daily life. Stroke, injuries, or neuromuscular disorders such as cerebral palsy can impair an individual’s ability to walk, thereby making activities of daily life much more difficult. Treatment of walking disorders targets muscles, but the mechanisms by which muscle forces influence body motion are poorly understood. Because important quantities like muscle forces cannot be measured during movement, experimental methods alone are insufficient for improving our understanding of the cause-effect relationships underlying the control of human walking. Muscle-driven simulations have emerged as powerful tools for calculating muscle forces and investigating how muscle forces influence body motion.

This dissertation presents a new “residual reduction algorithm” for improving the consistency in a model between motion and force data collected during motion capture, enabling the generation of simulations of long-duration gait movements such as walking and running. This algorithm has enabled hundreds of researchers around the world to generate simulations of walking. This algorithm was used to generate a thoroughly tested, three-dimensional muscle-driven simulation of an unimpaired man walking for ten gait cycles. This simulation has been made freely available at https://simtk.org/home/muscleprops so other researchers can download, reproduce, modify, and analyze this simulation to investigate questions regarding human walking without having to generate their own simulations from their own motion capture data, thereby saving thousands of man-hours of time for the field of computational biomechanics.

We used this simulation to resolve a long-standing mystery in movement science:​ how the human body responds to external disturbances when the central nervous system’s responses are substantially delayed. We showed that intrinsic properties of muscles help stabilize walking by responding instantaneously to disturbances, thereby complementing the central nervous system’s delayed response. This study demonstrates the utility of simulations for investigating questions about human walking that cannot be explored with experimental methods alone.

The residual reduction algorithm also enabled the generation of 32 simulations of eight subjects walking at four different speeds. We analyzed these simulations to determine which muscle groups make the largest contributions to mediolateral ground reaction force across a range of speeds. We showed that walking speed affects peak lateral ground reaction force in early stance and peak medial ground reaction force during early single support. The hip abductors are the largest contributors of medial ground reaction force at all walking speeds. The calf muscles, knee extensors, and adductors oppose the abductors by contributing lateral ground reaction forces.

The work presented in this dissertation has enabled several studies to be performed using muscle-driven simulations to gain insight into how muscles modulate walking and running movements. This dissertation presents two such studies that reveal insight into how muscle properties help stabilize walking and how muscles contribute to mediolateral stability during walking.