Gait analysis has been developed to quantify the mechanical aspects of human walking and includes computation of walking speed, joint kinematics, joint moments and so on. However, experimental gait analysis alone cannot provide enough information about muscle activities and how muscles interact with each other to move the body in a controlled fashion. Computer modeling and simulations have increased the applicability of gait analysis such that individual muscle forces and function can be predicted by the model. However, the complexity of a dynamic model depends on several factors, for example, the dimension of the model and individual muscle parameters.

This dissertation examined the sensitivity of dynamic models to several parameters and provided scientific guidelines for how to determine the complexity of a dynamic model. Specifically, this study compared muscle function in a sagittal plane model and a three dimensional model that simulated the same normal walking data. The sensitivity of predicted muscle forces to muscle parameters was also investigated based on a dynamic model with generic muscle parameters. Results showed that muscles that cross the ankle and knee joints had similar muscle activities in 2D and 3D models. Muscles which cross the hip joints had different muscle function between models. In addition, muscles had different sensitivity values in response to perturbation. Generally hip muscles and knee flexors were not sensitive to any perturbation. Ankle plantarflexors and knee extensors were sensitive to the values of tendon slack length and optimal fiber length.

Few studies have used dynamic models to directly simulate pathological walking data. This dissertation applied a forward dynamic model to simulate hemiparetic gait following stroke. Muscle functions were analyzed to determine the muscle compensation strategies that stroke subjects used. Experimental joint kinematics were successfully reproduced by the dynamic model. Ankle muscles and hip extensors were found responsible for the abnormal gait deviation.

The findings of this dissertation illustrate the importance of specifying model dimensionality and muscle properties when building dynamic models for human walking. The application of dynamic models to stroke gait have extended the clinical usage of computer simulation and revealed muscle function which experimental analysis cannot detect.