Training with robot-applied forces and virtual reality technologies has potential to restore motor functions in stroke rehabilitation, yet the optimal dosage of robotic therapy remains unclear. A fundamental principle in the motor learning literature is that neuroplasticity can be stimulated using distortions to the visual or mechanical environment where exaggerated mistakes (augmented error-feedback or EA) can accelerate and increase learning. However, robotic training using the error-augmentation methods have achieved only modest gains over conventional therapy. One explanation could be that stroke patients exhibit heterogeneous motor deficits and thus require more focused and individualized robotic rehabilitation. We propose a novel model-based approach that explores joint human-system performance by modeling dynamics for error-guided motor learning and recommends the optimal dosage schedule of EA to best enhance motor skill acquisition and retention. This proposal explores computational models of how various forms of EA affect motor learning and evaluates how the optimal EA dosage could be customized to healthy individuals and to chronic stroke patients. Research indicates that training specificity (i.e., training on the tasks you actually want to improve) is important to recovery of activities of daily living (ADL). Here we also test whether customized visual feedback based on optimal EA dosage can recover functional skills in 3D for stroke survivors. This work utilizes a data-driven modeling method in order to control and customize learning of novel motor skills using robotic technologies. This proposal will contribute to neuroscience knowledge by uncovering relationships between motor learning and error-feedback modulation. This proposal will improve engineering methods to optimize motor skill acquisition for the users operating intelligent-machines such as robotic tools, robot guided neuro-rehabilitation and prosthetics/assistive technologies.
This study examines how computational models of error-guided motor learning can be used to deliver optimal dosage of robot-based therapy in stroke rehabilitation. We investigate whether an optimal dosage therapy customized to an individual?s learning behavior and motor deficits can enhance acquisition of motor skills. This proposal will broaden neuroscience knowledge by uncovering computational mechanisms of motor learning and improve engineering methods to enhance human interaction with intelligent-machines such as robotic tools and prosthetics/assistive technologies.
