Human-like Robotic Controllers for Enhanced Motor Learning

The investigators will look at how the task performance and motor performance of individuals in dyadic physical interactions are affected by 1) different behavioral interaction conditions (i.e., solo task, collaboration task, competition task, or cooperation task); 2) the haptic impedance or stiffness of the virtual connection between dyadic peers (i.e., hard connection, medium connection, or soft connection); and 3) the skill level of the other partner (i.e., novice or expert). The investigators will be using both an ankle robot (M1 device) and a bilateral lower limb exoskeleton (H3/X2 device), and will collect EMG and EEG data. For Experiment A , the investigators will recruit healthy volunteers (n = 180) to work in dyadic pairs. With the collected data, the investigators will model how humans adapt force and impedance and share roles/specialize during various dyadic interaction behaviors, and use this knowledge to develop robot controllers that mimic movement error and force adaptation for enhanced motor performance. For Experiment B , the investigators will recruit healthy volunteers (n = 260), participants post-stroke (n = 88) and participants post-SCI (n = 88) to work in dyadic pairs within each population. The investigators will test the robot controllers following the models for mechanical adaptation and role sharing strategies between peers based on Experiment A. The investigators will also monitor single-joint and multi-joint movement error and force adaptation in regards to enhanced motor performance. The investigators will assess if the robot controllers can pass a "haptic Turing Test", rendering them indistinguishable with respect to human peers. A structural MRI will be obtained to be used for EEG source analysis. For Experiment C, the investigators will showcase the robot controllers by interfacing with participants post-stroke (n = 4) and participants post-SCI (n = 4) with the single-joint and multi-joint assistive robots to observe motor learning and functional outcomes with 10 training sessions per robot.