This paper deals with a design problem of Non-minimum phase systems with input constraints and unknown disturbance using the adaptive Robust Anti-windup Generalized Predictive Control (RAGPC) scheme. In essence non-minimum phase systems are characterized by presence of dead time and invariably has unstable inverse. This effect is not desirable in control processes especially when problems of static output feedback stabilization are concerned. Consequently, robust closed-loop stability of non-minimum processes with input constraint and disturbances becomes difficult to achieve. Addressing this issue provides a good motivation for the current study. Here, we present a new scheme for controlling such systems and in overall achieving a desired robust performance. To apply the proposed design scheme to a practical system, the system has been identified adaptively and thus the control scheme is applied. The practical system considered here is a Water level experimental system whose parameters are identified adaptively offline. Simulation results are presented to support the efficacy of the proposed scheme in controlling this kind of system.