Abstarct: One of the difficulties of model-baxsed control approaches is the uncertainty in system modeling. Therefore, estimation and compensation of uncertainty can improve the performance of control system. In this thesis, three novel adaptive approaches for estimation and compensation of uncertainty in the robust control of robot manipulator are proposed. In the first approach, Taylor series and in the second one Neural networks are used for uncertainty estimation and compensation using voltage control strategy instead of conventional torque control approaches. The third method is model-free and uncertainty is estimated by radial-basis-function networks. As advantages, the proposed control methods do not require proper selection of uncertainty bound parameters or bounding functions. As a result, one of the most important issues in designing the conventional robust control of robot manipulator which is the proper choice of uncertainty bounds and bounding functions are solved. In the conventional robust control, too low estimation of the uncertainty bounds may cause a higher tracking error while too high estimation of the bounds may cause saturation of input, higher frequency of chattering in the switching control laws and thus a bad behavior of the whole system. Scara robotic arm equipped with an electric permanent magnet motor is used for tracking control. In this research, kinematics and dynamics of robot and robust control methods equipped with the adaptive uncertainty estimators are introduced. In all of these proposed methods, closed-loop system stability is proved. To evaluate the performance of the proposed control system, controllers are simulated by MATLAB software. Simulation results show the efficiency of the control schemes.