The dynamic model of a robotic system is prone to parametric and structural uncertainties, as well as dynamic disturbances, such as dissipative forces, input noise and vibrations, to name a few. In addition, it is conventional to access only a part of the state, such that, when just the joint positions are available, the use of an observer, or a differentiator, is required. Besides, it has been demonstrated that some disturbances are not necessarily differentiable in any integer-order sense, requiring for a physically realizable but robust controller to face them. In order to enforce a stable tracking in the case of nondifferentiable disturbances, and accessing just to the robot configuration, an output feedback controller is proposed, which is continuous and induces the convergence of the system state into a stable integral error manifold, by means of a fractional-order reaching dynamics. Simulation and experimental studies are conducted to show the reliability of the proposed scheme.