This paper presents a sliding mode controller based on a sliding mode perturbation observer (SMPO) for the tracking control of robot manipulators. The SMPO is designed to estimate unmeasurable states and system perturbations that involve system non-linearities, disturbances, and unmodelled dynamics. The use of an SMPO allows the linearization of the input/output and the decoupling of the strongly coupled non-linear robot manipulator system merely by the feedback of joint angles, waiving the need for expensive tachometers. The controller design does not need an accurate model of the robot manipulator. Simulation studies are undertaken based on a two-link robot manipulator to evaluate the proposed approach. The simulation results show that the SMPO has more superior tracking control performance, with a wide range of payloads, in comparison with a sliding mode controller designed based on state feedback linearization with full states available.