Post-stroke motor disorders in the upper extremities, including hemiparesis, can inhibit daily activities. Rehabilitation in patients with upper extremity disorders can improve motor skills and restore motoric function to increase the patient's quality of life. The dynamics of patient movement during the rehabilitation process affect the movement of the DC motor as an actuator on the exoskeleton. The primary purpose of this paper is to enhance the robotic joint's position control performance by applying adaptive PID (Proportional-Integral-Derivative) control with online parameter tuning. The adaptation process utilizes feedback from positional error values and uses the sliding surface method, which is well-known for adapting to systems that experience external disturbances. Implementation of this adaptive PID controlled by a sliding surface verified by direct experiment on an exoskeleton prototype. An experiment involving six subjects who mimic a hemiparesis person in giving some disturbance to the system, which moves in the flexion-extension pattern. The results show that adaptive control can be applied to the exoskeleton prototype and give the best performance in frequency 0.014 Hz and 0.1 adaptation constant. The stability analysis shows that an exoskeleton with adaptive control can follow local asymptotic stability with position observer.