Design of a Powered Ankle Exoskeleton Controller With Adjustable Rise and Fall Time Assistance and Evaluation of Human Perception of Peak Timing

Abstract

Powered ankle exoskeletons are usually evaluated by how much they reduce metabolic cost, while how the assistance is perceived by the user remains largely unmeasured. This thesis pairs a broad account of the current state of exoskeleton control with a direct measurement of that experience. A systematic review of 143 studies, conducted under PRISMA guidelines, found that the field has moved toward learning-based control, with most designs pairing a classical controller with a computational algorithm that generally outperforms proportional–integral–derivative (PID) and other classical control methods alone, yet this work rests on a narrow evidence base and rarely measures the wearer's perception of assistance. To address this gap, a controller was developed for a bilateral Dephy EB-60 ankle exoskeleton. The controller reproduces the four-parameter ankle push-off torque curve introduced by Collins et al. (2015) and Zhang et al. (2017) which delivers a plantarflexion torque curve defined by its peak magnitude, start time and rise and fall time. The controller is adapted to the rigid-chain transmission of the Dephy EB-60 exoskeleton and allows for the adjustment of these parameters, recomputing the shape of the torque profile smoothly while a user walks. Using this platform, a psychophysical experiment is conducted examining how sensitive wearers are to changes in the timing of the timing of peak assistive torque. Twelve participants walked on a treadmill using the powdered exoskeleton while an adaptive staircase procedure moved the peak of the torque curve earlier or later to estimate smallest shift they could reliably feel, known as the just-noticeable difference. On average, participants noticed a shift of 4.10 ± 1.47% of stride and each wearer's individual sensitivity varied roughly four-fold. The same timing shifts systematically reshaped the exoskeleton ankle angle during push-off, driving the ankle further into plantarflexion as the assistance was delayed. This result shows how precisely peak timing needs to be controlled, giving future controller designs a concrete, human-centered threshold for perception.