抄録
A rehabilitation exercises are classified into a passive, an assistive, and an active exercise according to a trainer-trainee force ratio. When above training are realized by a device, it is necessary to regulate a generated force from a device based on trainee force. In order to apply Master-slave training device which has been developed to the above exercise trainings, a control system that can regulate a generated force from the slave device according to progress of training is proposed. If the trainee can not generate enough force to move an elbow, the device can assist the generated force to follow the reference angle from master device. To estimate a progress of a trainee, Neural Network is introduced to change stiffness in a compliance control system. In this paper, the proposed control method is described, and then effective using this control method is verified experimentally.
