Data-Driven Gait Detection for Single-Sensor Ankle Exoskeleton Control

Authors:

Hassan Farah, Jae-Ryeong Choi, Patrick Slade

Date Created:

2025-01-01

Course Title:

Professor:

Not specified

About Paper:

Walking is among the most energy-intensive activities in daily trained on data collected from treadmill experiments that varied life, which can be significantly taxing to those with mobility speed and incline, using inertial measurement units to provide impairments due to age, disability, or neuromuscular disease. labeled ground truth on heel strike occurrence. Preliminary results Various robotic devices have been developed for mobility yielded an accuracy of ~93% and a precision of ~92% . We plan to assistance, including lightweight ankle exoskeletons worn on the conduct further validation of the exoskeleton on various walking lower leg that have been shown to reduce the metabolic cost of conditions to assess muscle activation upon assistance. Our results walking. These devices provide assistive plantarflexion torque via demonstrate the efficacy of accurately detecting gait events using a an electromechanical clutch that engages during the heel strike single embedded sensor, enabling a significantly simplified design phase of the gait cycle. However, conventional exoskeletons for the ankle exoskeleton. Our discovery has minimized the need require external sensors on the body to detect heel strikes, which for external sensors and hardware while maintaining performance, complicates hardware and limits usability. Here we present a meaning the exoskeleton can assist using data based solely on quasi-passive ankle exoskeleton that detects heel strikes using a its mechanical structure. This enables easier replication and machine learning model trained solely on data from a single sensor deployment,contributingtowardstranslationofankleexoskeletons embedded within the clutch. A classification tree model was from laboratory environments to real-world, everyday use. Harvard-Amgen Scholars Program 5

Abstract:

Walking is among the most energy-intensive activities in daily trained on data collected from treadmill experiments that varied life, which can be significantly taxing to those with mobility speed and incline, using inertial measurement units to provide impairments due to age, disability, or neuromuscular disease. labeled ground truth on heel strike occurrence. Preliminary results Various robotic devices have been developed for mobility yielded an accuracy of ~93% and a precision of ~92% . We plan to assistance, including lightweight ankle exoskeletons worn on the conduct further validation of the exoskeleton on various walking lower leg that have been shown to reduce the metabolic cost of conditions to assess muscle activation upon assistance. Our results walking. These devices provide assistive plantarflexion torque via demonstrate the efficacy of accurately detecting gait events using a an electromechanical clutch that engages during the heel strike single embedded sensor, enabling a significantly simplified design phase of the gait cycle. However, conventional exoskeletons for the ankle exoskeleton. Our discovery has minimized the need require external sensors on the body to detect heel strikes, which for external sensors and hardware while maintaining performance, complicates hardware and limits usability. Here we present a meaning the exoskeleton can assist using data based solely on quasi-passive ankle exoskeleton that detects heel strikes using a its mechanical structure. This enables easier replication and machine learning model trained solely on data from a single sensor deployment,contributingtowardstranslationofankleexoskeletons embedded within the clutch. A classification tree model was from laboratory environments to real-world, everyday use. Harvard-Amgen Scholars Program 5

Source:

Harvard / Prairie View A&M University | Biology | 2026 / 2025

Topics:

exoskeleton, ankle, using, heel, strike, sensor, gait, single, walking, trained, significantly, mobility

Co-authors:

@hassanfarah107 , @jaeryeongchoi108 , @patrickslade109