Development of A Multi-Axial Transtibial Powered Prosthesis Driven by EMG Signals

Abstract

Human locomotion extends beyond walking at self-selected speeds. It encompasses running, walking over varied terrain, being able to move laterally and other such non-cyclic behaviour, all while being able to maintain one's stability. The loss of a limb necessitates adaptation. In the case of lower limb amputation, and depending on the level of amputation, it could mean having to re-learn how to walk. The objective of this research was to develop a control strategy that was primarily driven by user acquired real-time electromyography (EMG) signals, and to test said control strategy on a developed transtibial powered prosthesis prototype. The development of the control strategy was based on analysing how able-bodied individuals maintained their dynamic stability during ambulation, particularly as they walked over fixed, uneven terrain. This was done by conducting a gait experiment with six able-bodied participants who performed walking trials on both level ground and a fixed, uneven terrain. The observed gait adaptations implemented by the participants included increased ground reaction force (GRF) variation, particularly along the frontal plane (foot eversion and inversion). These were coupled with aligned co-activation of the participants' lower limb antagonist muscle pairs. These findings highlighted the importance of the human ankle being able to move along multiple planes and how this movement, along with synchronised activation of the lower limb muscles, facilitated the maintenance of dynamic stability. In the pursuit of developing a volitionally controlled transtibial powered prostheses, various control approaches were tested. The EMG data acquired during the gait experiment formed the basis for the explored controllers. The control approach that yielded the best accuracy when tested using both offline and real-time EMG data was then implemented as the final control strategy to be tested on a prosthesis prototype. Validating the developed control strategy meant testing its functionality with real-life application; to this end, a multi-axial transtibial powered prostheses prototype was developed. The prototype was designed in such a way that it could mimic able-bodied gait both in form, i.e. range of motion (ROM), and in function, i.e. output torque during the gait cycle. The chosen control strategy was implemented on the prototype and tested in real-time with able-bodied participants walking over level ground and on the same fixed, uneven terrain used for the gait experiment. The control strategy was able to generalise to new participants, there was only a 10% decrease in accuracy using new data compared to using the training data. Testing the prototype, and by extension the control strategy, led to range of motion (ROM) and torque output results that were similar to able-bodied ankle ROM and torque output. The peak torque was observed around push-off (powered plantarflexion) which demonstrated the prototypes ability to supply energy at appropriate stages of the gait cycle. The findings indicated that the developed control strategy could enable traversal over level-ground and fixed, uneven terrain based solely on real-time user EMG data. It was also found that the control strategy could facilitate movements that, unlike human walking, were not cyclic in nature. Therefore, the aim of this research project was achieved.

Date of Award	31 Dec 2018
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Zhenmin Zou (Supervisor) & Lei Ren (Supervisor)