Hybrid Force/EMG Based Control System for Rehabilitation Robot

Abstract

This paper presents a hybrid control system of rehabilitation robot with force and EMG signals. The proposed control system is implemented on the elbow joint of the 4 DOF universal exoskeleton. Admittance control method is applied to control this rehabilitation robot. However, the transient response of the admittance control cloud lead in a large overshoot when the user moves exoskeleton joint quickly then suddenly stops. Hence, the EMG sensor is used to detect the muscle contraction and then the force input will be set to zero for improving transient response of the hybrid controller. Furthermore, the generalized regression neural network (GRNN) is applied for predicting the static gravity force compensation. The experimental result indicates that the GRNN can predict the static gravity force with accuracy of 97.32%. Moreover, 83.13% of the transient response is improved by the utilization of the EMG signal in the hybrid controller.