MORE | Spring 2020

Learning Deep Neural Interaction Policy for ExoSkeleton Control

Data icon, disabled. Four grey bars arranged like a vertical bar chart.

The purpose of this research is to develop a deep-learning-based approach to control a custom-built hip exoskeleton. A key component is to leverage learning-based techniques to adapt across different walking behaviors among multiple users. At the core, learning from demonstration (LfD) is used to extract the inherent motion pattern from the user demonstrations during training.  Based on these patterns the model proactively assists or resist the user during their motion. This allows the model to encourage a healthy posture by resisting or supporting the user by applying assistive force. Future work will focus on creating better predictive models and cater better to the needs of the user.

Student researcher

Dhaivat Dholakiya

Engineering (robotics)

Hometown: Tempe, Arizona, United States

Graduation date: Spring 2020