Activity-Invariant Human Augmentation

To augment human mobility in daily life, wearable robot controllers must generalize beyond controlled walking environments conventionally studied in the lab – driving the need for activity-invariant exoskeleton controllers. This research project aims to introduce an exoskeleton controller able to modulate assistance to reduce human joint effort during a wide array of activities (e.g., walking, lunging, and jumping) using deep learning. The first goal of this project is to develop an opensource dataset of human lower-limb biomechanics during a variety of natural human movements. Using the resulting dataset, the second goal of this project focuses on the optimization of a deep learning-based exoskeleton controller, able to generalize with changes in human activity. The resulting exoskeleton system is expected to bridge the gap between the successes of in-lab exoskeleton technology and the needs for improved human mobility in the real world.

Lab members:

Dean Molinaro
Keaton Scherpereel
Ethan Schonhaut
Justine Powell

Collaborators:

Dr. Max Shepherd
Google X