Abstarct: Due to the increasing importance of robots in today's life, the use of robots with light weight, small size, higher speed, more accuracy and greater power is required. Since the rigid and heavy robots are not effective to perform well in advanced applications, it is very desirable to use flexibile robotic manipulators. In the meantime, in order to produce high torque in low speed using transmission system is inventible. The major root for flexibility is using harmonic drive in the transmission system. By joint flexibility we mean that the motor position and joint angle are not related through a simple gain but instead there is a dynamic relation between them. This can be modeled in the simplest way as a torsional spring . On the other hand, joint flexibility makes the system under actuated in which the number of control inputs is less than the number of degree of freedoms (DOFs). So one cannot use each input to control a related DOF. These problems have been challenged many researchers through the last two decades and yet it is an open problem. This thesis is devoted to the type-2 fuzzy control of flexible joint robots by considering the stability problem and actuator saturation. Moreover, a systematic design methodology is provided to demonstrate how this proposed method can be extended to robots with multi degree of freedom while still ensuring that stability is guaranteed. To overcome the complexity of the robot manipulators dynamics, we use voltage control strategy. Compared to torque control strategy, voltage control is simpler, less computational and more efficient because of not using the robot model. The implementation of torque control strategy has two major drawbacks. First, torque control laws are inherently involved in complexity of the manipulator dynamics characterized by nonlinearity, largeness of model, coupling, uncertainty and joint flexibility. Second, actuator dynamics may be excluded from the controller design. Another issue addressed in this thesis is to investigate the applicability and capability of type-2 fuzzy system in the estimation of uncertainties. As a new approach, adaptive type-2 fuzzy system is successfully used as an uncertainty estimator in robotic system. A problem that has not been addressed so far in this area. It is verified that the proposed adaptive type-2 fuzzy system can model the uncertainty as a nonlinear function of the joint position error and its time derivative. The proposed adaptive type-2 fuzzy system has an advantage that does not employ all system states to estimate the uncertainty. Since uncertainty may include parametric uncertainty, unmodeled dynamics and external disturbances, control system is capable of overcoming the uncertainties in the wide range The design includes two interior loops: the inner loop controls the motor position using the proposed adaptive type-2 fuzzy control while the outer loop controls the joint angle of the robot using an adaptive proportional-integral-derivative (PID) controller. By defining proper sliding condition using gradient descent there PID control gains are updated. Furthermore, to better examine the performance of the type-2 fuzzy controller, a type-1 fuzzy controller is also designed and their performances in the face of unmodeled dynamics are compared. Simulation results show that type-2 fuzzy control can be an effective way to overcome the uncertainties in the wide range.