Title: Motion Prediction and Actuator Design for Augmenting Exoskeletons
Speaker: Bon H. (Brandon) Koo, Human Systems Lab
Abstract: Demand for a streamline, lightweight, and most importantly intuitive exoskeleton has increased recently due to advancements in the defense, industrial, and aerospace sectors. However, currently available, or recently developed, active exoskeletons lack deployment potential due to limitations in fluency; it is often the case that the exoskeletons currently touted fail to intuitively interact with a user. In this presentation, two angles to solving this problem of low fluency are proposed. First, it may be beneficial for future exoskeleton platforms to predict a user’s motion: a control mode largely unseen in driving powered exoskeletons today. Second, the failure in recreating human motion accurately may come from the nature of actuators used in most exoskeletons developed for augmenting purposes; utilizing biomimetic actuators instead that can simulate higher order muscle dynamics without complex control methods may increase fluency. Through both simulations as well as prototype testing, both angles show promise. This study simultaneously demonstrates 1) the feasibility of machine learning and generational categorization algorithms’ potential to predict human motion and 2) the capability of artificial muscle to not only faithfully recreate simple muscle contraction but also to simulate more complex muscle phenomenon. Combining advancements in these two thrusts may yield active augmenting exoskeletons featuring higher fluency.