Human-in-the-Loop Optimization of the Stiffness and Alignment of a Prosthetic Foot to Reduce the Metabolic Cost of Walking

Abstract

Improper tuning of prosthetic foot properties to the individual user limits the efficacy of current state-of-the-art prosthetic feet in terms of walking economy. This study aims to explore the potential of human-in-the-loop optimization to individually optimize prosthetic foot stiffness and alignment to decrease the metabolic cost of walking of transtibial amputees. 10 transtibial amputees underwent an optimization protocol while walking on a treadmill with an experimental prosthetic foot with tuneable stiffness and alignment. We aimed to minimize the metabolic cost of walking by optimizing the stiffness and alignment of the prosthetic foot, using an evolutionary optimization algorithm. The metabolic cost of walking during the post-test using optimal settings was compared with the pre-test using standard settings, and the post-test using standard settings. Human-in-the-loop optimization of the tuneable prosthetic foot resulted in optimal stiffness ( $4.41~\pm ~0.17$ Nm/±) and alignment ( $2.40~\pm ~0.97^{\circ }\text {)}$ settings that differ between participants. Walking on the prosthetic foot with optimized settings during the post-test resulted in a significant reduction in metabolic cost compared to the pre-test with standard settings (−10.6%). The metabolic cost during the post-test with standard settings was in between the pre-test with standard settings (−6.6%) and the post-test with optimal settings (−4.3%), indicating that part of the decrease in cost could be explained by motor adaptation of the user. Human-in-the-loop optimization can individually tune the stiffness and alignment of a prosthetic foot to lower the metabolic cost of walking for transtibial amputees and provides different optimal settings for each individual participant. Both optimization of prosthetic components and motor adaptation of the user contributed to the reduction in metabolic cost, which corroborates that human-in-the-loop optimization could enhance the efficacy of prosthetic devices.