Multiple model switched repetitive control for FES-based tremor suppression

Abstract

Tremor is a rhythmic, approximately periodic oscillation of the limbs caused by a range of neurological disorders. Functional electrical stimulation (FES) can help reduce tremor by artificially stimulating opposing muscles, thereby decreasing the oscillation's amplitude. However, various traditional control methods have not proven effective. Repetitive control (RC) has the potential to completely suppress tremor, however previous applications of repetitive controllers have limitations. In particular, they heavily rely on an accurate model of the limb dynamics, and their effectiveness declines steeply due to factors like muscle fatigue, spasticity, and modeling inaccuracies.

This thesis first reviews conventional tremor suppression technologies, before summarising existing applications of RC to suppress tremor. Then RC theory is summarised with focus on a general class of linear algorithms. Following this, their robustness is analysed for the first time using the gap metric. This enables a robust stability margin to be developed that characterises robustness of the system to unstructured modelling uncertainty. Building on these results, a design procedure for RC is proposed to maximise its robust performance. A control structure with an additional feedback compensator is also proposed and a corresponding robust design procedure is developed for this more general case. Both design procedures are applied to a model of the tremulous wrist in simulation. The results show that the repetitive controllers used in previous research are not robust to model uncertainty and have performance limitations. Finally, a multiple model switched repetitive control (MMSRC) framework is established, which has the potential to solve the current limitations of existing RC approaches. Experimental validation is performed with four participants, showing that MMSRC effectively suppresses tremor even in the presence of severe modeling uncertainty and fatigue, unlike conventional RC methods which often become unstable under these conditions. This is an important step towards home-based tremor suppression where model identification and expert tuning are not possible.

More information

Identifiers

ORCID for Chris Freeman: orcid.org/0000-0003-0305-9246

ORCID for Ann-Marie Hughes: orcid.org/0000-0002-3958-8206

Catalogue record

Export record