World health organization reports, annually more than 15 million people worldwide suffer a stroke and cardiovascular disease, among which 85% of stroke patients incur acute arm impairment, and 40% of victims are chronically impaired or permanently disabled. This results a burden on the families, communities and to the country as well. Rehabilitation programs are the main way to promote functional recovery in these individuals. Since the number of such cases is constantly growing and that the duration of treatment is long, an intelligent robot could significantly contribute to the success of these programs. We therefore developed a new 5DoFs robotic exoskeleton named MARSE-5 (motion assistive robotic-exoskeleton for superior extremity) that supposed to be worn on the lateral side of upper arm to rehabilitate and ease the shoulder, elbow and forearm movements. This paper focused on the design, modeling, development and control of the proposed MARSE-5. To control the exoskeleton, a nonlinear sliding mode control (SMC) technique was employed. In experiments, trajectory tracking that corresponds to typical passive rehabilitation exercises was carried out. Experimental results reveal that the controller is able to maneuver the MARSE-5 efficiently to track the desired trajectories.