Bayesian Optimisation of Exoskeleton Design Parameters

Abstract

Exoskeletons are currently being developed and used as effective tools for rehabilitation. The ideal location and design of exoskeleton attachment points can vary due to factors such as the physical dimensions of the wearer, which muscles or joints are targeted for rehabilitation or assistance, or the presence of joint misalignment between the human subject and exoskeleton device. In this paper, we propose an approach for identifying the ideal exoskeleton cuff locations based on a human-in-the-Ioop optimisation process, and present an empirical validation of our method. The muscle activity of a subject was measured while walking with assistance from the XoR exoskeleton (ATR, Japan) over a range of cuff configurations. A Bayesian optimisation process was implemented and tested to identify the optimal configuration of the XoR cuffs which minimised the measured EMG activity. Using this process, the optimal design parameters for the XoR were identified more efficiently than via linear search.