Georgia Institute of Technology
Enabling Personalized Prosthetic Control Using Self-Learning and Bayesian Optimization
Abstract
dc:description.abstractLower-limb powered prostheses have the potential to restore missing limb function and replicate able-bodied biomechanics. However, state-of-the-art prosthetic control systems often rely on user-specific offline tuning and extensive data collection, requiring new users to endure lengthy multi-hour sessions. This process is impractical and significantly limits the widespread adoption of powered prostheses. In this work, I enable prosthetic controllers to continually learn user-specific gait patterns and improve performance in real-time, eliminating the need for offline data collection. I demonstrate that substantial improvements can be achieved within minutes of walking, resulting in prosthetic assistance that closely mimics able-bodied biomechanics across various terrains. To ensure safe, user-agnostic performance out of the box, I propose data-efficient methods that leverage open-source able-bodied datasets to develop generalized controllers for prosthetic applications. When deployed in real-time, these methods incrementally enhance a prosthetic device’s ability to deliver precise, biomimetic assistance. Recognizing the diversity in user preferences, physiology, and biomechanics, I introduce auto-tuning algorithms that adapt control parameters to optimize gait symmetry measures, achieving personalized objectives for each user. These innovations mark a significant step toward accessible, safe, and personalized prosthetic control, paving the way for the broader adoption of powered prostheses.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Mechanical Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Maldonado-Contreras, Jairo Yobani
- Advisor dc:contributor.advisor
-
- Young, Aaron
- Committee members dc:contributor.committeemember
-
- Chang, Young-Hui
- Zhao, Ye
- Gombolay, Matthew
- Fey, Nick
Subjects
dc:subject × 7Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/81566