Aaron Young
Georgia Institute of Technology, GA·Biomedical Engineering·Early-career
Walking is tied to independence, mood and health, and people recovering from stroke, living with an amputation, or simply aging often lose the ability to get around their own neighborhoods. Powered exoskeletons and robotic prostheses can push at the hip, knee or ankle to help, but they face two stubborn problems: the assistance has to be hand-tuned for each person, and it usually works for only one activity, typically steady treadmill walking.
- Fields
Biomedical Engineering
Engineering
Medicine & Clinical Research
- Topics
exoskeleton
wearable robotics
robotic prosthesis
- Grants
- 4 known grants, through Jan 2029
What they work on
- Compare personalized hip versus ankle exoskeleton assistance for improving walking on both legs after stroke, testing the hypothesis that hip assistance yields larger walking-speed gains because stroke survivors compensate at the hip for ankle weakness.
- Test whether a wearable overground biofeedback system that cues better paretic-leg positioning works synergistically with hip and ankle exoskeletons to improve gait mechanics, forward propulsion of the center of mass, and walking speed.
- Evaluate clinically relevant outcomes, including walking speed, gait energetics and biomechanics, for stroke survivors walking in real-world environments with and without the optimized exoskeleton and biofeedback systems.
- Build deep learning models that estimate a wearer's underlying lower-limb joint efforts from wearable sensor data, and use these internal state estimates to drive exoskeleton assistance across new users and new tasks.
- Collect natural human movement data from exoskeleton wearers across cyclic walking and discrete tasks such as cutting, jumping and stepping over obstacles, to train and test task-agnostic controllers.
- Develop ankle exoskeleton control that maps wearable sensor readings directly to assistance across complex daily tasks such as shuffling, bending to lift objects and entering a vehicle, without subject-specific tuning, and open-source the datasets and models.
- Create a self-adaptive intent recognition system that learns an individual patient's gait patterns, and translate it to a robotic knee/ankle prosthesis for clinical testing in people with transfemoral amputation.
- Formulate a human-in-the-loop actor-critic reinforcement learning framework that maximizes a multi-objective reward to self-adapt control policy across users and environments, validated on a hip exoskeleton with stroke survivors.
More about the research
Research here attacks both problems with artificial intelligence trained on large amounts of human movement data, aiming for controllers that learn an individual's gait and generalize across the messy variety of real-world tasks. Four projects run from foundational algorithms through device-agnostic ankle control to clinical testing with stroke survivors and people with above-knee amputation.
Why it matters
Stroke survivors and people with above-knee amputation often cannot walk far enough or fast enough to take part in community life, and current wearable robots must be tuned by experts for one activity at a time. Controllers that learn a person's own gait and adapt across real-world tasks could make exoskeletons and prostheses usable outside the lab, potentially increasing walking speed, stability and independence. Open datasets and models would also let other robotics groups build on the work.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Personalized Hip vs Ankle Exoskeleton Assistance with Biofeedback to Enhance Mobility Post-Stroke
NIH · Apr 2024 – Jan 2029
FRR: A new strategy for task-agnostic control of robotic exoskeletons by estimating underlying biological effort using deep learning
NSF · Oct 2023 – Sep 2027
Collaborative Research: Towards Task-Agnostic and Device-Agnostic Ankle Exoskeleton Control for Mobility Enhancement
NSF · Sep 2023 – Aug 2027
Closest labs
Myunghee Kim
University of Illinois Chicago, IL·Biomedical Engineering·Early-career
Build and evaluate a personalized robotic prosthesis for people with below-knee amputations that combines adaptive optimization with biofeedback so assistance adjusts in real time to each user's movement patterns.
Tonya Stone
Mississippi State University, MS·Engineering
Design an elastic spring for an ankle foot prosthesis using machine learning-based optimization to determine the curvatures and contours that maximize energy return during walking.
Zhishan Guo
North Carolina State University, NC·Engineering
Develop a gait capture method using only wearable inertial motion unit (IMU) sensors, using knowledge distillation so a multi-sensor teacher network trains a simpler IMU-only student network to predict walking in daily life.
The school
Georgia Institute of Technology: 50 labs on PI Pages, $1.5B a year on research, 572 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.