Dataset for Task-Agnostic Exoskeleton Control via Biological Joint Moment Estimation
No Thumbnail Available
Author(s)
Molinaro, Dean
Scherpereel, Keaton
Schonhaut, Ethan
Evangelopoulos, Georgios
Shepherd, Max
Advisor(s)
Editor(s)
Collections
Supplementary to:
Abstract
Lower-limb exoskeletons have the potential to transform the way we move, but current state-of-the-art controllers cannot accommodate the rich set of possible human behaviors which range from cyclic and predictable to transitory and unstructured. We introduce a task-agnostic controller that assists the user based on instantaneous estimates of lower-limb biological joint moments from a deep neural network. By estimating both hip and knee moments in-the-loop, our approach provided multi-joint assistance through our autonomous, clothing-integrated exoskeleton. When deployed during 28 activities, spanning cyclic locomotion to unstructured tasks (e.g., passive meandering and high-speed lateral cutting), the network accurately estimated hip and knee moments with an average R2 of 0.83 relative to ground-truth. Our approach significantly outperformed a best-case task classifier based on splines and impedance parameters. When tested on 10 activities (including level walking, running, lifting a 25lb weight, and lunging), our controller significantly reduced user energetics (metabolic cost or lower-limb biological joint work depending on the task) relative to the Zero Torque condition, ranging from 5.3-19.7%, without any manual controller modifications among activities. Thus, this task-agnostic controller can enable exoskeletons to aid users across a broad spectrum of human activities, a necessity for real-world viability.
Sponsor
Date
2024-08
Extent
Resource Type
Dataset
Resource Subtype
Rights Statement
Unless otherwise noted, all materials are protected under U.S. Copyright Law and all rights are reserved