Kinematic-Synergy-Based Robotic Feedback Promotes Short-Term Retention of Altered Human Bimanual Motor Behavior

Lasting improved motor performance is a major goal of human-centered robotics research. Literature shows adaptation to robotic intervention but points to washout of adapted motor patterns and after-effects. Retention would affirm desired effects of robotic intervention on motor behavior. We suggest the form of robotic intervention, like dependency and output characteristics of feedback, significantly influences retention. One form of robotic intervention, proven to retain after-effects, is error-enhancement, which has exclusively been used in simple reaching or tracking tasks. A method for multi-objective, bimanual tasks-a focus of applications, like rehabilitation, surgical training, and occupational exoskeletons-would be valuable. In this work, we propose a kinematic-synergy-based torque feedback to improve short-term retention of adapted motor patterns in multi-objective, bimanual tasks while coupled with a robotic system. Synergies are motor pattern elements, theoretically comprising both simple and complex movements. We tested retention of alterations of individuals' primary kinematic synergy for two multi-objective, bimanual tasks performed with an exoskeleton. Half of participants received kinematic synergy-dependent torque feedback, and half received a velocity-dependent force field. We hypothesized 1) kinematic synergies of complex tasks adapt to augmented system dynamics and 2) retention of altered primary kinematic synergies can be achieved through synergy-dependent torque feedback. Hypotheses were tested on intra-subject Pearson's correlation coefficient of primary synergies across experimental sets. Results indicate robotically augmented system dynamics indeed induce kinematic synergy adaptation ( ${p}\lt {0}.{05}$ ), but washout occurs when reverted. However, targeting synergies permits at least short-term retention ( ${p}={0}.{04}$ ), which has better implications for long-term motor training.