Abstarct: In this dissertation, control of completely non-affine pure-feedback systems in the presence of uncertainties is investigated. First, the adaptive control of parameterized pure-feedback systems is considered. These parametric uncertainties include both linear and nonlinear parameterization and belong to an unknown compact set, i.e., no prior knowledge is required on the bound of the unknown parameters. Therefore, the class of systems considered here is much more general than the systems in previous related works. For non-linearly connected parameters term, using parameter separation technique, the bounding function is obtained which is linear in new unknown parameter. For linearly connected parameters term, a positive function of parameter is applied instead of using the parameter itself. By employing the two mentioned techniques directly in combination of the backstepping and time scale separation procedures, the virtual/actual control inputs are defined as solutions of fast dynamic equations. In this approach, the adaptation law of unknown parameters can be derived baxsed on Lyapunov theory in the backstepping technique and there is no need to design state predictor for this purpose. Therefore, it results in higher accuracy and avoids complexity. The new theorem in singular perturbation theory is presented for closed loop stability of these systems. Second, the robust control of these systems in the presence of matched and unmatched uncertainties are considered. Using singular perturbation theory, high gain filters are designed to estimate these uncertainties. By combination of the backstepping and time scale separation, the virtual/actual control inputs are obtained and the fast variables arising from filters, are employed to cancel the effect of the uncertainties in control design. One of common assumptions in previous related works is that bound of these uncertainties is known. However, in this approach this restrictive assumption is removed. Finally, we present the output feedback control problem for completely non-affine pure-feedback systems with immeasurable states because it is usually impossible that all the states are available in the actual process. First, the state observer is designed to estimate the immeasurable states. Then, the time-scale separation concept and the backstepping technique are combined to develop the approximate version of virtual/actual control laws. This technique has eliminated the problem of “explosion of complexity” caused by the traditional backstepping approach and the circular design problem which exists in pure-feedback systems. The simulation results for different pure-feedback systems are provided to validate the effectiveness of the proposed approach.