Emerging robotic exoskeletons have the potential to enhance performance of non-disabled users, restore normative leg biomechanics to impaired users, or rehabilitate limb function after neurological disorders. However, these devices have not been widely adopted due to their rigidity, bulkiness, acoustic noise, cost, battery life, and/or non-intuitive control. While parallel industries are providing powerful, lightweight, and inexpensive motors with long-lasting batteries, the exoskeleton research community is now largely focused on mechanical design, interface, and control challenges.
One body of work has shown that human walking effort can be reduced by customizing the assistance torque pattern to the user for a given activity through human-in-the-loop optimization. While some results have extended outside the lab, it is unclear how to quickly generalize these methods across many users and activities. Another body of work has shown that task-agnostic control methods can handle a continuous range of activities without user-specific calibration, for example, predicting human joint torques across activities to reduce muscle effort or metabolic cost.
However, it is unclear how these versatile control methods can be personalized for optimal assistance, especially for clinical patients. This workshop will explore specificity vs. generalizability across users and activities with the goal of finding benefits from both perspectives. This theme will be discussed as it relates to both control and device hardware optimization.
