EVALUATION COMMITTEE

The Faculty Council has appointed the following adjudication committee to evaluate the thesis and the associated lecture: 

• Prof. Silvestro Micera, EPFL, Lausanne, Switzerland
• Prof. Prof Jacqueline S Hebert, University of Alberta, Canada
• Dr. Sabata Gervasio, HST, Aalborg University, Denmark (Chairman).

Moderator:
Prof. Strahinja Dosen, Health Science and Technology, Aalborg University

ABSTRACT

Prosthetic technologies aim to restore lost functionalities of individuals with a limb difference and improve their quality of life. User-prosthesis interfaces are at the heart of this technology. They convert the user’s intentions to prosthesis movements and provide feedback about the state of the prosthesis to the user, creating closed-loop bidirectional interactions. These interfaces thereby determine the level of skill with which a user can control their device. Despite the growing efforts in designing these (closed loop) interfaces, it is unknown how they affect the acquisition of skilled prosthesis usage and moreover how different interfaces compare in this respect.

In this thesis, we leverage the phenomenon of speed accuracy tradeoffs – the scientific equivalent of ‘haste makes waste’ – and develop an experimental framework to evaluate and understand how different closed-loop user-prosthesis interfaces affect user skill. In a series of experiments, participants were instructed to use the prosthesis to grasp and lift objects in a ‘timed force-matching task’ with the objective of applying a specified force on the objects within a specified time, using two different interfaces. We then investigated how the speed-accuracy tradeoffs afforded by each interface differed and how they evolved across days by combining performance level metrics such as accuracy of achieving a certain force with behavioral outcomes that measure how variable and smooth the user-generated commands were. Notably, by characterizing the interfaces across a range of speeds and accuracies, we demonstrated how movement speed critically affected user skill – a facet neglected by current experimental frameworks. Finally, we also developed parametric models of the tradeoff to estimate user skill.

Taken together, we have developed an experimental and model-based framework which we believe can be used to gain a fundamental understanding of skill learning in prosthesis control and to standardize the evaluation of interfaces in various task contexts.