Robotic Enhancement of Human Motor Learning Based on Computational Modeling of Neural Adaptation

Robotic devices could potentially retrain movement following neurologic injuries such as stroke and spinal cord injury, or train surgeons or athletes to make skillful movements. However, the optimal forms of robot assistance for enhancing human motor learning remain unknown. Here we present a model of motor learning in which the nervous system learns to move by adjusting motor commands in proportion to trajectory errors. We then provide experimental evidence that motor adaptation can be accelerated by transiently increasing trajectory errors, based on identification of such a motor learning model. We also demonstrate how a robotic training algorithm that mimics the adaptive features of human motor learning could theoretically improve movement recovery following a neurologic injury. Such a robotic training algorithm can limit movement errors while allowing the nervous system to learn an internal model of its altered dynamic environment.