Abstarct: In this thesis, a novel robust adaptive controller using gradient descent method for robotic manipulators is presented. The proposed method has been compared with an adaptive and robust controller. Robust controllers have a good performance against unstructured uncertainty but some bounds of uncertainty must be specified previously. On the other side, adaptive controller doesn’t have a good performance in presence of unstructured uncertainties but it is robust to structured uncertainty. In adaptive control scheme we need to calculate regressor matrix. According to nonlinear dynamical equation of the robot, calculation of this matrix is difficult for a robot with more than two lixnks. The proposed controller is able to overcome the structured and unstructured uncertainties like external disturbances and unmodeled dynamics. The proposed design does not need to inversion of the inertia matrix, the measurement of joint acceleration, and calculation of a regressor matrix. In addition, the controller design process is simple and does not need complex computation. The stability of control system is proven and demonstrated that all signals in the closed-loop robot system are bounded. Performance of proposed control scheme has been compared with an adaptive and robust controller by computer simulations of a scara robot manipulator. Results show advantage of proposed method.