ABSTRACT
The present paper develops a virtual tennis system using robotic devices for the rehabilitation of upper arm movements, and discusses an evaluation method of human motor skills for natural stroke in the dynamic task. In the training, a trainee manipulates the handle of an impedance-controlled robot to hit a ball by a racket toward the center of a circular target on the wall with considerations of motion smoothness and timing. A reference hand motion for the target task is clarified through a set of training experiments with 6 health volunteers and computationally expressed in the framework of a minimum jerk model. The designed index based on the reference model will be effective to quantitatively evaluate the recovery of motor skills for natural stroke in the virtual tennis task.
