TY  - THES
AU  - Garcia Rosas, Ricardo
Y2  - 2020/11/19
Y1  - 2020
UR  - http://hdl.handle.net/11343/251820
AB  - Upper-limb loss affects over 541,000 people in the US, and over 3,500 new amputations are reported each year in countries like Italy and the UK. The daily life of people living with upper-limb loss is severely impacted as the arm is a human's principal means of interaction with the environment. Moreover, the limitations of current human-prosthetic interfaces result in prosthesis users relying on compensatory motion to achieve activities of daily living, which may result in overuse injuries. This is due to prosthesis users only being able to control the degrees-of-freedom in the prosthesis sequentially. To address this challenge, the prosthetics community has looked into motion-based human prosthetic interfaces.

Novel motion-based human-prosthetic interfaces use the motion of the residual-limb to determine the motion of the prosthesis. Typically, this relationship between the residual-limb and prosthesis is established from the motion of able-bodied individuals. However, their application to prosthesis users has been a challenge due to individual differences in motor behaviour and amputation physiology. Therefore, it has been identified in the literature that kinematic synergy-based HPIs need to be personalised to their users.

The scope of the research presented in this thesis is to provide a framework for autonomously personalising human-prosthetic interfaces. The proposed framework is based on a data-driven optimisation approach. The contributions of this thesis surrounding the proposed data-driven-based framework are as follows. First, the feasibility of using online optimisation methods in motion-based human-prosthetic interfaces is demonstrated experimentally. Second, the features of motor preference and motor adaptation in human motor behaviour, which affect the performance of a task with a motion-based prosthesis, are experimentally observed and characterised in a grey-box model. Third, an online personalisation algorithm for human-prosthetic interfaces was developed based on the algorithm of Fast Extremum Seeking. The algorithm uses the grey-box model of human motor preference and adaptation to inform the design of the components of the algorithm.
    
An alternative model-based method for motion-based human-prosthesis interface personalisation is also proposed, where user-specific kinematic information is employed. This novel ``task-space synergy'' incorporates task information in the formulation of kinematic synergy-based human-prosthetic interfaces. The method uses desired hand path information, a kinematic model of the human-prosthesis arm, and the motion of the residual-limb to determine the motion of the prosthesis joints.
KW  - Human-robot interaction
KW  - Prosthetics
KW  - Online optimisation
KW  - Extremum seeking
KW  - Human motor behaviour
KW  - Shared control
T1  - Online Personalisation of Human-Prosthetic Interfaces
L1  - /bitstream/handle/11343/251820/ef0afae5-f36e-ea11-94bb-0050568d7800_Garcia-Rosas_Thesis_2020.pdf?sequence=1&isAllowed=y
ER  -