Walking is a complex task and requires the coordination of many muscles. After stroke, muscle coordination is impaired and may compromise walking performance. The objective of this dissertation was to develop and apply simulation tools to quantify the role of muscles during slow gait in healthy individuals and in post-stroke hemiparesis.

Because individuals with post-stroke hemiparesis often exhibit equinus gait, which is commonly associated with excessive ankle plantarflexion at foot strike and knee hyperextension later during stance, a muscle-actuated forward dynamic simulation of self-selected speed gait was perturbed to emulate this gait deviation. We found that excess plantarflexion at initial foot placement can by itself induce considerable knee extension in midstance and this effect was further pronounced when intrinsic muscle responses were suppressed.

To facilitate the development of muscle-actuated forward dynamic simulations of slow and pathological gait, a portable parallel version of a simulated annealing optimization algorithm was designed. The linear scalability and robustness of the optimization algorithm was verified for a simple parabolic function and for a more complex forward dynamic simulation of pedaling on up to 32 processors. The parallel implementation of the simulated annealing algorithm significantly reduced time to convergence, thus making computationally expensive problems in biomechanics tractable.

Using this optimization algorithm, a muscle-actuated forward dynamic simulation of healthy slow gait was generated based on the experimental kinematics and kinetics of healthy adults walking at an extremely slow speed (0.3 m/s). Individual muscle forces in midstance were perturbed and the effect on the height of the body center of mass and on hip, knee, and ankle extension assessed. Similar perturbations were applied to a simulation of self-selected speed gait (1.5 m/s). We found that uniarticular ankle extensors (e.g., soleus) provided less midstance support of the body in slow gait and the uniarticular knee extensors (e.g., vastii) and the biarticular calf muscles (e.g., gastrocnemii) more.

A muscle-actuated forward dynamic simulation of post-stroke hemiparetic gait was produced that emulated the experimental kinematics and the vertical ground reaction force of a single stroke survivor. The contributions of paretic and non-paretic muscles to midstance support were compared with muscle contributions in healthy slow gait. Differences in muscle excitations and midstance body configuration caused paretic and non-paretic ankle plantarflexors to contribute less to midstance support than in healthy slow gait. Paretic knee and hip extensors compensated for stance phase co-activation of paretic ankle dorsiflexors and knee flexors which opposed support.

This dissertation presents the first simulations of healthy slow and post-stroke hemiparetic gait and the corresponding contributions of individual muscles to body support. Future studies will use enhanced modeling and simulation techniques coupled with experimental data to suggest potential therapeutic interventions targeting improved walking performance in stroke survivors.