There is an increasing need for transfemoral prostheses that provide gait support, stability, safety and comfort. Although there are many prostheses available in different levels of complexity and price, there is still room for improvement. It has been proved that the cost of transport (CoT) for walking is significantly increased for transfemoral amputees with respect to their healthy peers. Assisting push-off is one of the main challenges in prosthesis design. Push-off is normally achieved by plantarflexion of the ankle joint. Prosthesis designs should aim to restore this function in order to lower the amount of energy needed for walking.

This study aims to investigate the effect of prosthesis design on the gait pattern through musculoskeletal modelling and predictive simulations. Two prosthesis designs are modelled for these purposes, after which several variations on these models are made. It is hypothesised that the prosthesis that assists in push-off through ankle plantarflexion, should result in a gait pattern that is closer to a healthy one. It should also decrease the CoT. Furthermore, we aim to evaluate the use of modelling and simulations in the customisation of prostheses.

OpenSim was used to create a total of eight models based on a model with 9 degrees of freedom and 18 muscles: a healthy person, a conventional prosthesis model, two scaled versions of the conventional prosthesis model, the walkMECH prosthesis and three variations on the walkMECH. SCONE was used to find an optimal gait pattern for each of the models through the CMA-ES method. CoT-, gait-, degrees of freedom- and reaction force objectives were minimised. The results were evaluated by comparing the CoT, joint angles, ground reaction forces and muscle activation of each model.

The CoT for the healthy model was found to be higher than reported before, based on both experimental and simulation studies. As a result, we have little confidence in the CoT estimation of our models. This is further exacerbated by the finding of a lower CoT for the conventional prosthesis than for the healthy model, in contrast to earlier reports. The results for most other measures were irregular, making it difficult to draw conclusions from them. It is expected that the predictive optimisations did not reach a global minimum, and that the results are therefore not accurate. Future research should aim to solve this problem. It should also be attempted to find the cause of the difference in CoT between our simulations and those of others.

No conclusions could be drawn from the results. Nonetheless, there is a clear potential for the use of musculoskeletal modelling and predictive simulation in the investigation of the effects of prosthesis design on gait.