Title: Neural Control of Exoskeletons
Speaker:  Hosea Siu, LL Technical Staff and HSL Alum

Zoom:  https://mit.zoom.us/j/96678010991

Exoskeletons that assist in human movement have the potential to greatly impact human performance in the military, industrial, and medical domains. However, existing controller designs only allow exoskeleton operation in a narrow range of speeds and terrains, and until recently, were not optimized for specific users. While current controllers can work well in steady-state movements, they do not perform as well when transitions - like from standing to running or sprinting to standing - are involved. Additionally, optimizing exoskeleton assistance for specific users is a process that involves many hours of human-in-the-loop controller tuning.

The Neural Control of Exoskeletons program aims to bring lower-body exoskeletons closer to operational status by advancing three major enablers. First, wearable sensors and embedded computing are used to predict human motions and joint torques in various types of locomotion and transitions through machine learning. Second, a model of muscle energy usage takes the torque predictions to optimize user-specific exoskeleton assistance across different kinds of locomotion without human-in-the-loop tuning. Finally, the effects of exoskeleton assistance on human balance and stability were studied to better understand exoskeleton impacts during tasks that require increased balance on the part of the user.

Apart from developing leg exoskeletons to expand the capabilities of healthy humans, the lessons learned from this program can also be applied to additional use cases in the clinical domain and more broadly for wearable robotics.