Human-in-the-loop Bayesian optimization of wearable device parameters

Myunghee Kim, Ye Ding, Philippe Malcolm, Jozefien Speeckaert, Christoper J. Siviy, Conor J. Walsh, Scott Kuindersma

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

The increasing capabilities of exoskeletons and powered prosthetics for walking assistance have paved the way for more sophisticated and individualized control strategies. In response to this opportunity, recent work on human-in-the-loop optimization has considered the problem of automatically tuning control parameters based on realtime physiological measurements. However, the common use of metabolic cost as a performance metric creates significant experimental challenges due to its long measurement times and low signal-to-noise ratio. We evaluate the use of Bayesian optimization - a family of sample-efficient, noise-tolerant, and global optimization methods - for quickly identifying near-optimal control parameters. To manage experimental complexity and provide comparisons against related work, we consider the task of minimizing metabolic cost by optimizing walking step frequencies in unaided human subjects. Compared to an existing approach based on gradient descent, Bayesian optimization identified a near-optimal step frequency with a faster time to convergence (12 minutes, p < 0.01), smaller inter-subject variability in convergence time (± 2 minutes, p < 0.01), and lower overall energy expenditure (p < 0.01).

Original languageEnglish (US)
Article numbere0184054
JournalPloS one
Volume12
Issue number9
DOIs
StatePublished - Sep 2017

Fingerprint

walking
employment opportunities
Equipment and Supplies
Walking
exoskeleton
system optimization
energy expenditure
Costs and Cost Analysis
Signal-To-Noise Ratio
Global optimization
Prosthetics
Time measurement
Energy Metabolism
Noise
Costs
Signal to noise ratio
Tuning
sampling

ASJC Scopus subject areas

  • Biochemistry, Genetics and Molecular Biology(all)
  • Agricultural and Biological Sciences(all)
  • General

Cite this

Kim, M., Ding, Y., Malcolm, P., Speeckaert, J., Siviy, C. J., Walsh, C. J., & Kuindersma, S. (2017). Human-in-the-loop Bayesian optimization of wearable device parameters. PloS one, 12(9), [e0184054]. https://doi.org/10.1371/journal.pone.0184054

Human-in-the-loop Bayesian optimization of wearable device parameters. / Kim, Myunghee; Ding, Ye; Malcolm, Philippe; Speeckaert, Jozefien; Siviy, Christoper J.; Walsh, Conor J.; Kuindersma, Scott.

In: PloS one, Vol. 12, No. 9, e0184054, 09.2017.

Research output: Contribution to journalArticle

Kim, M, Ding, Y, Malcolm, P, Speeckaert, J, Siviy, CJ, Walsh, CJ & Kuindersma, S 2017, 'Human-in-the-loop Bayesian optimization of wearable device parameters', PloS one, vol. 12, no. 9, e0184054. https://doi.org/10.1371/journal.pone.0184054
Kim, Myunghee ; Ding, Ye ; Malcolm, Philippe ; Speeckaert, Jozefien ; Siviy, Christoper J. ; Walsh, Conor J. ; Kuindersma, Scott. / Human-in-the-loop Bayesian optimization of wearable device parameters. In: PloS one. 2017 ; Vol. 12, No. 9.
@article{c984c869497a49fa9bacb0ecd052ec0a,
title = "Human-in-the-loop Bayesian optimization of wearable device parameters",
abstract = "The increasing capabilities of exoskeletons and powered prosthetics for walking assistance have paved the way for more sophisticated and individualized control strategies. In response to this opportunity, recent work on human-in-the-loop optimization has considered the problem of automatically tuning control parameters based on realtime physiological measurements. However, the common use of metabolic cost as a performance metric creates significant experimental challenges due to its long measurement times and low signal-to-noise ratio. We evaluate the use of Bayesian optimization - a family of sample-efficient, noise-tolerant, and global optimization methods - for quickly identifying near-optimal control parameters. To manage experimental complexity and provide comparisons against related work, we consider the task of minimizing metabolic cost by optimizing walking step frequencies in unaided human subjects. Compared to an existing approach based on gradient descent, Bayesian optimization identified a near-optimal step frequency with a faster time to convergence (12 minutes, p < 0.01), smaller inter-subject variability in convergence time (± 2 minutes, p < 0.01), and lower overall energy expenditure (p < 0.01).",
author = "Myunghee Kim and Ye Ding and Philippe Malcolm and Jozefien Speeckaert and Siviy, {Christoper J.} and Walsh, {Conor J.} and Scott Kuindersma",
year = "2017",
month = "9",
doi = "10.1371/journal.pone.0184054",
language = "English (US)",
volume = "12",
journal = "PLoS One",
issn = "1932-6203",
publisher = "Public Library of Science",
number = "9",

}

TY - JOUR

T1 - Human-in-the-loop Bayesian optimization of wearable device parameters

AU - Kim, Myunghee

AU - Ding, Ye

AU - Malcolm, Philippe

AU - Speeckaert, Jozefien

AU - Siviy, Christoper J.

AU - Walsh, Conor J.

AU - Kuindersma, Scott

PY - 2017/9

Y1 - 2017/9

N2 - The increasing capabilities of exoskeletons and powered prosthetics for walking assistance have paved the way for more sophisticated and individualized control strategies. In response to this opportunity, recent work on human-in-the-loop optimization has considered the problem of automatically tuning control parameters based on realtime physiological measurements. However, the common use of metabolic cost as a performance metric creates significant experimental challenges due to its long measurement times and low signal-to-noise ratio. We evaluate the use of Bayesian optimization - a family of sample-efficient, noise-tolerant, and global optimization methods - for quickly identifying near-optimal control parameters. To manage experimental complexity and provide comparisons against related work, we consider the task of minimizing metabolic cost by optimizing walking step frequencies in unaided human subjects. Compared to an existing approach based on gradient descent, Bayesian optimization identified a near-optimal step frequency with a faster time to convergence (12 minutes, p < 0.01), smaller inter-subject variability in convergence time (± 2 minutes, p < 0.01), and lower overall energy expenditure (p < 0.01).

AB - The increasing capabilities of exoskeletons and powered prosthetics for walking assistance have paved the way for more sophisticated and individualized control strategies. In response to this opportunity, recent work on human-in-the-loop optimization has considered the problem of automatically tuning control parameters based on realtime physiological measurements. However, the common use of metabolic cost as a performance metric creates significant experimental challenges due to its long measurement times and low signal-to-noise ratio. We evaluate the use of Bayesian optimization - a family of sample-efficient, noise-tolerant, and global optimization methods - for quickly identifying near-optimal control parameters. To manage experimental complexity and provide comparisons against related work, we consider the task of minimizing metabolic cost by optimizing walking step frequencies in unaided human subjects. Compared to an existing approach based on gradient descent, Bayesian optimization identified a near-optimal step frequency with a faster time to convergence (12 minutes, p < 0.01), smaller inter-subject variability in convergence time (± 2 minutes, p < 0.01), and lower overall energy expenditure (p < 0.01).

UR - http://www.scopus.com/inward/record.url?scp=85029604628&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85029604628&partnerID=8YFLogxK

U2 - 10.1371/journal.pone.0184054

DO - 10.1371/journal.pone.0184054

M3 - Article

C2 - 28926613

AN - SCOPUS:85029604628

VL - 12

JO - PLoS One

JF - PLoS One

SN - 1932-6203

IS - 9

M1 - e0184054

ER -