The applications of robotics and automation technology to the therapies of neuromuscular and orthopedic impairments have received increasing attention due to their promising prospects. In this paper, we present an actuated upper extremity exoskeleton aimed to facilitate the rehabilitation training of the disable patients. A modified sliding mode control strategy incorporating a proportional-integral-derivative sliding surface and a fuzzy hitting control law is developed to ensure robust and optimal position control performance. Dynamic modeling of the exoskeleton as well as the human arm is presented and then applied to the development of the fuzzy sliding mode control algorithm. A theoretical proof of the stability and convergence of the closed-loop system is presented using the Lyapunov theorem. Three typical real-time position control experiments are conducted with the aim of evaluating the effectiveness of the proposed control scheme. The performances of the fuzzy sliding mode control algorithm are compared to those of conventional proportional-integral-derivative controller and conventional sliding mode control algorithm. The experimental results indicate that the position control with fuzzy sliding mode control algorithm has a bandwidth about 4 Hz during operation. Furthermore, this control approach can guarantee the best control performances in term of tracking accuracy, response speed, and robustness against external disturbances.