Data Driven Intent Recognition Framework

A critical aspect of exoskeleton control that has to date introduced a performance limitation is the ability of the exoskeleton to recognize the intent of the operator so it can apply assistance to their desired motion. This intent recognition effort is typically solved using ad-hoc methods where subject matter experts make design decisions and tune transitions to identify intended maneuvers as reliably, quickly and robustly as possible. This approach however, is limited by the capacity of the expert designers and is not readily optimized, extended, or validated. With this effort, we will build a salable intent recognition framework that allows for reliable development of intent recognition algorithms using data driven techniques to allow for quantitative design decisions and performance optimization that can scale across almost any hardware system.