A Simple Model of Skill Acquisition in a Dynamic Balance Task

Abstract

We are interested in modeling human motor skill acquisition in complex dynamic tasks. Understanding how humans learn to do a task can unveil new directions for learning of control in robotics. Modeling motions of human learners can also help us in generating effective motions for graphics and entertainment robotics. Human motor adaptation has been widely studied in structured tasks such as goal directed arm movements, to answer questions such as ‘what is learned’ and ‘how is it learned’ [1]. However, it is unclear whether the mod- els proposed using studies of such structured tasks apply to highly dynamic motor skills such as balancing and locomotion. We study motor skill acquisition in a motion capture friendly, highly dynamic task of balancing on a bongo-board.