Shape memory alloys are a group of smart materials which are of a lot interest in different fields such as robotics. Shape memory alloys have a complex electro-thermo-mechanical dynamic and due to which their modeling is difficult. In fact the main problem for the modeling is that their temperature and stress change simultaneously. So choosing a proper model which can predict shape memory alloys’ behavior in presence of temperature-stress dependency is so important.Nonlinear hysteresis and uncertainties makes shape memory alloys actuators’ control difficult. Modeling of shape memory alloys in a way to be used in control algorithm, choosing a proper control method to compensate for hysteresis and overcome the uncertainties problem as well as kinetics of the robot finger and eventually studying the results are the main parts of this research. In this study for investigating the shape memory alloys behavior and controllers’ performance, a rotary manipulator which is actuated with shape memory alloys is modeled. Also changing the temperature of the SMAs is done by joule heating. For the rotary manipulator mentioned some control algorithms are designed. First a position-based PID controller is designed, and it is found that this controller cannot perform well for all the desired angular positions (set-points). Also the performance of the controller is not good for tracking. Regarding that the temperature and transformation of shape memory alloys depends on the stress, choosing the stress as the set-point can be so efficient. So a sliding mode controller based on the stress of the SMA wire is designed. The results show better performance for this controller than the PID one for both regulation and tracking. By studying the results, it is found that choosing different values for the slope of sliding surface have an effect on the performance of the controller. So to have a better performance a fuzzy tuner is designed for tuning the slope of sliding surface. The position error is the input for the fuzzy tuner and the slope of sliding surface is the output. The better performance of the controller than the previously designed controllers is showed. Finally to study the robustness of the fuzzy-sliding mode controller the performance of the controller is investigated by considering the moment of inertia, damping coefficient and heat convection coefficient as the uncertainties. The performance of the proposed controllers is studied for both a 1DOF and a 3-link rotary manipulator which are actuated with shape memory alloys. The results show that the sliding mode controller cannot perform well for all set-points for the 3-link manipulator, but the fuzzy-sliding mode controller does have a good performance for all set-points. Key words: hape memory alloys, sliding mode controller, fuzzy tuner, Robustness