Most manipulator motion controllers require joint velocity feedback. Whenever joint velocities are not measurable, they are estimated from the joint positions. However, velocity estimates tend to be inaccurate under low-speed motion or low sensor resolution conditions. Moreover, velocity estimators may either be susceptible to model uncertainties or introduce additional dynamics (e.g., phase lag) to the control loop. Consequently, direct substitution of velocity estimates into the controller results in the deterioration of the control performance and robustness margin. Therefore, this paper proposes a robust position-feedback motion controller which gets rid of the problems of uncompensated dynamics and model uncertainties introduced by velocity estimators. Furthermore, a globally asymptotically stable system, which is robust with respective to model parameter variations, is guaranteed. Theoretical analysis and experimental verifications are carried out. The results demonstrate that the proposed controller is robust and outperforms the conventional computed torque plus proportional integral differential (PID) controller.