December 4, 2023
By Ian Sargent

A collaboration between Georgia Tech professors Aaron Young and Matthew Gombolay has secured a four-year National Science Foundation (NSF) grant. The grant targets the development of a novel approach to controlling orthotic exoskeletons, which can help restore mobility for patients with lower limb impairments.

“This is an awesome opportunity for the lab to advance the use of data-driven machine learning techniques to provide intuitive and biomechanically appropriate control for assistive exoskeleton technology,” Young said.
 

Interdisciplinary Collaboration at Georgia Tech

Young, an associate professor in the George W. Woodruff School of Mechanical Engineering, is the director of the Exoskeleton & Prosthetic Intelligent Controls (EPIC) Lab which focuses its research on the development and enhancement of powered control systems for orthotic and prosthetic devices.

Gombolay is an assistant professor in the School of Interactive Computing and is the director and founder of the Cognitive Optimization and Relational (CORE) Robotics Laboratory. CORE specializes in developing algorithmic techniques that can enable robotic systems to learn through experience rather than relying on a rigid program or remote user.

Gombolay, emphasized the interdisciplinary nature of the group's work, saying, “Robotics is highly interdisciplinary and systems driven. The pace of research and innovation is progressing so quickly that individual researchers and schools must specialize, which means no single school can do everything. Working across schools opens up opportunities and resources to innovate.”
 

A New Approach for Real-World Impact

The two plan to leverage the power of deep learning to create a more comprehensive method of controlling powered orthotic devices. Current control methods typically rely on understanding the user’s environment and applying different control strategies based on different environments and tasks.

“This means a large number of controllers needs to be developed to provide a real-world control system,” Young explained. “Instead, we focus on a unified system that is constantly estimating the internal state of the human and does not require specific information about the external environment. We believe this will be a much more scalable approach for real-world deployment and use of exoskeletons outside of the lab.”
 

The Challenge of Data Integration

Young and his team have amassed substantial data related to human locomotion, converting movements like cutting, dancing, and playing tug-of-war into information. This data trains a deep learning system, co-developed by Young and Gombolay, enabling the system to recognize actions and their relationship to human physiology. This recognition will eventually help guide exoskeletons to offer tailored support for a variety of tasks.

Sophisticated computer programs are needed to make use of the data, which can come in different forms and formats. “My role on the project is to develop new machine learning algorithms to enable exoskeletons to do a better job leveraging freely-available data to more efficiently and adapt to individual users and transitions among walking tasks – autonomously and from the wearer’s feedback –  reducing the effort required to manually-tune and configure these devices” Gombolay said.

Such transitions, like going from level ground to climbing stairs, can happen quickly, and a control unit that can adapt to these transitions while also providing a personalized level of assistance has the potential to vastly improve the efficiency of an orthotic exoskeleton.

“This is my first grant addressing exoskeletons,” Gombolay added, “and I am very excited to work with Young.”