| Summary: | Mathematical models of human movement have the potential to provide information about how and why humans move the way they do. However, due to the existing gap between the current simple and complex models of running, it is unclear as to the level of complexity required to model normal distance running. Therefore, the aim of this PhD was to develop an “appropriately complex” biomechanical model of running. The first investigation aimed to assess the validity of the fundamental assumption of the spring mass model, the simplest model of running. The model assumes that during running the human body acts similar to a point mass bouncing on a massless linear spring of constant stiffness. It was found that forefoot strikers exhibited a relatively linear force-length relationship, but rearfoot strikes did not. In addition, significant differences were found when a range of definitions for calculating lower limb stiffness were compared to the most physically consistent definition. A series of models, each with increasing complexity, were developed using OpenSim. Simulations were compared to experimental data and agreement appeared to increase as model complexity increased. Due to problems with using a fixed length segment, introducing a knee joint alone prevented successful simulations across the entire stance phase. In contrast, a model with passive knee and ankle joints resulted in good agreement between the experimental and simulated center of mass (CoM) trajectories. However, joint kinematics and ground reaction forces (GRFs) did not show as good a match. The next level of complexity was to add actuation at the ankle joint. Within OpenSim this required introducing a contact model at the foot-ground interface, and led to difficulties in ensuring sensible behaviour of the contact model as well as matching the CoM trajectories. Including actuation provided better agreement between the joint kinematics and GRFs. Therefore, it is predicted that if these difficulties can be overcome, a three-segment sagittal plane model with torsional springs at the knee and ankle, and a small amount of actuation, will be able to match experimental data to within the measurement error of the data.
|
|---|
