IPAB Workshop - 18/01/2018

Speaker

Daniel Gordon

Title

Human-in-the-loop Optimisation of Exoskeleton Design

Abstract

Increasingly, exoskeletons and other assistive robotic devices have been used to great effect for rehabilitation and limb replacement. Poorly fitting exoskeletons apply force in the wrong places and can cause discomfort or work against the intended aim of the device. As such, the optimal scenario is for exoskeletons to be custom-designed for a patient. This is not always possible, however, due to cost and time constraints. Adjustable exoskeletons, which have attachment points whose position can be modified to suit a specific user, can offer a cheaper alternative to custom-made exoskeletons in the setting of a rehabilitation centre, where large numbers of patients may need to be treated.

One difficulty with the use of adjustable exoskeletons is determining the optimal locations for the attachment points. These can vary between two subjects due to differences in physical characteristics (e.g. limb lengths, heights, muscle mass) and differences in their rehabilitation goal (e.g. whether the aim is to target the hamstrings or gastrocnemius muscles). If adjustable exoskeletons are to be used effectively in a clinical setting, a means of efficiently identifying the optimal cuff positions for an individual is required. This talk will discuss a method for doing this based on using Bayesian optimisation as part of a human-in-the-loop optimisation process.