Bayesian optimization improves control of exoskeletons for comfort and energy use
Multi-Objective Human-in-the-Loop Bayesian Optimization of a Lower-Limb Exoskeleton
RoboticsHuman-Computer Interaction
Summary
Assistive devices like exoskeletons need to work well for each individual, since everyone moves differently and prefers different things. The authors show a new way to tune such devices using feedback from the person wearing it, balancing multiple goals like comfort and energy efficiency. Their method finds many good settings that suit different preferences rather than just one fixed balance. They tested this on a lower-limb exoskeleton and showed their approach finds solutions people prefer. They also provide open-source software to help others try it on wearable devices.
What this means in practice
- •For assistive device engineers: Customize wearable exoskeleton controls to balance comfort and energy efficiency for individual users across changing conditions.
- •For rehabilitation technology teams: Deploy personalized multi-objective tuning methods for lower-limb assistive devices to improve patient comfort and reduce fatigue.
Authors
Neil Janwani, Matthew T. Lerner, Aaron J. Young, Maegan Tucker
Abstract
Human-in-the-loop optimization (HILO) is a common approach for optimizing the control of assistive devices to account for the wearer's unique biomechanics and subjective preferences. However, despite research suggesting that a person may have a different prioritization of objectives depending on time-varying factors such as the environment, their mood, or energy levels, existing HILO approaches only consider a single objective or enforce a fixed weighting on a set of objectives. Neither approach is capable of representing an individual's preferences over objectives. In this work, we propose Multi-Objective Human-in-the-loop Bayesian Optimization (MO-HILBO), which builds on explicit multi-objective Bayesian optimization to efficiently infer a personalized set of Pareto-optimal controllers. We compare our approach with an existing multi-objective HILO method and experimentally demonstrate MO-HILBO on a lower-limb exoskeleton across two objectives: metabolic cost (efficiency) and ordinal human feedback (comfort). We find that MO-HILBO (1) discovers Pareto-optimal controllers, and (2) that the pairwise ordering of points on the Pareto front itself is consistent with validation trials. Lastly, we open-source mohilo, a Python package for running both HILO and MO-HILBO on wearable devices: https://dynamicmobility.github.io/mohilo/.