The desire for higher performance from the structure and mechanical specifications of robot manipulators has spurred designers to come up with flexible joint robots (FJR). Several new applications such as space manipulators and articulated hands necessitate using FJRs. This necessity has emerged new control strategies required, since the traditional controllers implemented on FJRs have failed in performance. Since 1980's many attempts have been made to encounter this problem and now, several methods have been proposed including various linear, nonlinear, robust, adaptive and intelligent controllers. Among these, only a few researchers have considered practical limitations such as actuator saturation in the controller synthesis, although it is a real practical drawback to achieve good performance.

On the other hand actuator saturation has been considered by the control community from early achievements of control engineering. During 50's and 60's at the beginning era of optimal control, researchers have been working on saturation, introducing bang-bang control methods. Over the last decade the control research community has shown a new interest in the study of the effects of saturation on the performance of systems. In fact it can be said that in the past, researchers were encountered actuator saturation as a drawback and they had developed methods to avoid it, while now researchers develop methods to achieve a desirable performance in the presence of actuator saturation encountered as a limitation.

A common classical remedy for systems with bounded control is to reduce the bandwidth of the control system such that saturation seldom occurs. This is a trivial weak solution, since even for small reference commands and disturbances the possible performance of the system is significantly degraded. This idea (reduction in bandwidth by reduction in the closed loop gain) is practical and "easy", hence it motivates some researchers to propose an "adaptive" reduction in bandwidth consistent with the actuation levels. The "adaptation" process is done under supervision of a supervisory loop, and as proposed in a paper, it can be accomplished through complex computations, which seem not to be practically implementable. In order to come up with an

online implementable controller in presence of saturation, we have proposed a fuzzy logic supervisory control in this work. In this topology, the fuzzy logic is set to be "out of the main loop", at a supervisory level, at the aim of preserving the essential properties of the main controller. This idea is then modified to use with composite controller for FJRs. It is observed in various simulations that by including this supervisory loop to the controller structure, the steady state performance of the system is preserved, and moreover, the stability of the overall system which may be affected by addition of a saturation block, will retain. In other words the supervisor can remove instability due to saturation. The stability analysis of the overall system, however, is essential for the closed loop structure for susceptible applications of the FJRs such as space robots. Robust stability proof for the "composite + supervisor" method has been also given in detail in this thesis.

A second approach has been also proposed in this thesis. The robust methods proposed in this approach are simpler than the previous method (composite + supervisor) in structure, and moreover, they need only the feedback of the link position. In this approach using a frequency weighted penalty function of the control action is recommended in the mixed sensitivity minimization. Furthermore, in order to decrease the amplitude of the control action while keeping the desired bandwidth, a mixed  minimization method has been also proposed. Numerical controller design has obtained using LMI method. Simulations verify the superior performance of the mixed  method compared to that of the composite PID+PD controller and the  mixed sensitivity controller.

In this thesis, practical implementations of the proposed methods have also accomplished to verify the theoretical results. These implementations illustrate the effectiveness of the proposed methods in practice.