Although many everyday manipulation tasks are performed using the hand and wrist, relatively few studies have focused on the neuromotor control of the wrist, especially during human-robot interaction. To address this gap in knowledge, we developed a wrist robot, the MR-SoftWrist, for use during functional magnetic resonance imaging (fMRI) to investigate how the brain learns to control movement during force perturbation tasks.
Behavioral models suggest that the brain learns to reduce performance errors resulting from force perturbations via neuromotor adaptation. To characterize the adaptive behavior of the wrist, I collected data on 48 healthy young adults performing a force perturbation task with the MR-SoftWrist. In our task, participants were cued to make alternating wrist rotations to move a cursor between targets displayed on a screen at (+/-10,0) degrees along the flexion-extension and radial ulnar deviation axes of the wrist. We evaluated typical wrist pointing behavior in a no force mode, and adaptation in a learnable curl-force mode, in which the robot applied a velocity-dependent torque proportional to the measured velocity. Behavioral analysis showed that while subjects undergo significant adaptation, they may also use alternative impedance control strategies to reject curl-force perturbations. Existing models of motor adaptation, necessary to localize neuromotor adaptation networks in the brain via fMRI, do not account for this behavior.
To validate these behavioral findings, we developed the UDiffWrist, a non-MR compatible robot to execute the same force perturbation task with less measurement noise than the MR-SoftWrist. Additionally, we incorporated EMG during task performance to measure muscle activation to assess impedance control via muscle co-contraction. This experiment (in progress) will be used to determine an appropriate model of motor control for investigation of neural activity in future tasks executed with the MR-SoftWrist during fMRI, to localize learning networks associated with neuromotor adaptation and alternative, impedance control networks.