Jelle Broekzitter performed his MSc thesis project under supervision of Dick Plettenburg and David Abbink
Abstract
Many users of body-powered upper extremity prostheses experience difficulties using their device and a large group abandons usage altogether. Shoulder control via Bowden cable is widely used because of the intuitive use and low cost, but requires large shoulder movements and high operating forces that, according to literature, often exceed the upper limit of around 20N that would allow for fatigue-free prolonged use. Implementing by-wire control reduces friction forces due to shorter cables, but it also allows for prosthesis control to be treated as a telemanipulation problem: workspace extension methods could prove effective in further reducing these movements and forces. To test if these methods are indeed applicable, three different modes of control – proportional gain control, non-linear variable gain control, and velocity control – were implemented in an ideal virtual environment. Performance was measured at both ends of the range of motion using an experiment based on Fitts’ translational tapping task. It was hypothesised that variable gain control would improve speed during gross positioning and accuracy during fine positioning, improving overall performance, while velocity control would perform worse. The results show that the hypothesis holds, as well as improved controller performance when using a variable gain. It can be concluded that variable gain control in shoulder actuated prostheses can be beneficial and it would be worth exploring in real-life applications.