Personalization of Exoskeleton Control

Every individual walks differently, and personalized exoskeleton control is essential, particularly for clinical populations with gait disabilities. Our goal is to tailor exo control to each user's specific needs while minimizing data requirements and experimental burden (rapid personalization). To achieve this, we focus on the following research threads:

• Online Continual Personalization: Developing exo control frameworks that adapt in real-time to a user's unique gait dynamics across diverse locomotor tasks.

• Sim-to-Real Personalization: Implementing zero- and few-shot personalization strategies utilizing control policies trained within predictive computer simulations.

• Generative Data Synthesis: Utilizing generative AI to synthesize personalized movement data, mitigating the need for extensive human data collection.

• Computer Vision-Based Exo Control: Leveraging exocentric and egocentric computer vision to capture user-specific movements and design environment-aware controllers.

Team Members

• Changseob Song

• Diya Denish

• Adrian Krieger

• Marvin Lim

• David Newsom

Past Projects