Abstarct: Utilizing a fuzzy system to approximate a function under the fuzzy universal approximation theorem conditions, is followed by the fuzzy approximation error. Without compensating the fuzzy approximation error, asymptotic tracking in adaptive fuzzy control of robot manipulators, is not achieved. In this thesis, the problem of fuzzy approximation error compensation in adaptive fuzzy control of robot manipulators, is stated and solutions are made to improve the tracking performance. It is mathematically proven that the fuzzy approximation error is compensated and the tracking error convergence to zero is achieved. In the proposed adaptive fuzzy controllers, the fuzzy approximation error is a part of the uncertainties of the control system. A robust term is added to the adaptive fuzzy controller to compensate the uncertainties, which also compensates the fuzzy approximation error. Employing an adaptive term instead of the robust term in adaptive fuzzy control, sliding mode adaptive fuzzy control and bi-level adaptive fuzzy control of rehabilitation robots, are proposed to compensate the fuzzy approximation error. Voltage control strategy is used instead of torque control strategy, in the proposed adaptive fuzzy controllers of electrically driven robot manipulators. In this strategy, the motor model is used instead of the robot model in the control system design and motor voltages instead of joint torques, are given to the robotic system as the inputs. Therefore, the proposed control systems are simpler, less computational and more applicable. The proposed adaptive fuzzy controllers are compared with adaptive fuzzy controllers with no fuzzy approximation error compensators. The proposed controllers are simulated on an electrically driven SCARA robot manipulator in presence of uncertainties and the results are evaluated through comparisons. Furthermore, bi-level adaptive fuzzy control is simulated on a Lower Limb Rehabilitation Robot. The results show that the precision of the tracking performance is improved.