Interactive motor learning with the autonomous training assistant: A case study

Abstract

At-home exercise programs have met limited success in rehabilitation and training. A primary cause for this is the lack of a trainer’s presence for feedback and guidance in the home. To create such an environment, we have developed a model for the representation of motor learning tasks and training protocols. We designed a toolkit based on this model, the Autonomous Training Assistant, which uses avatar interaction and real-time multi-modal feedback to guide at-home exercise. As an initial case study, we evaluate a component of our system on a child with Cerebral Palsy and his martial arts trainer through three simple motion activities, demonstrating the effectiveness of the model in representing the trainer’s exercise program.