Abstarct: In this thesis, sliding mode control method is studied for controlling robot manipulator because of its robustness against model uncertainties and external disturbances, and also its ability in controlling nonlinear and MIMO systems. In this method, using high control gain to overcome uncertainties lead to occur chattering phenomena in control law which can excite unmodeled dynamics and maybe harm the plant. Additionally, in use of this control methodology, we must have good knowledge about the model of the system which is generally difficult affair. Different approaches, such as intelligent methods, are used to abate these drawbacks. In order to enhancement the sliding mode controller performance, we have used fuzzy logic and neural networks. For this purpose, we have proposed and designed three controllers, here. In the first one, we have used a PID outer loop in the control law then the gains of the sliding term and PID term are tuned on-line by a fuzzy system, so the chattering is avoided here. In the second, two practical issues such as boundedness of control torque magnitude and existence of joint frictions are considered and we apply two stable sliding mode and fuzzy controllers, cooperatively such that, each one alleviates other one drawbacks and improves control performance. In the third, we approximate the terms which associate to model in sliding mode controller design by using of a neural network and like the first case, the fuzzy gain tuning is applied, here. In all of three above cases, presented simulation results confirm the above claims and demonstrate the performance improvement in each case.