This paper presents a new model reference robust adaptive controller which does not require a priori knowledge of the norm desired controller parameter vector. The algorithm is designed according to a reduced order plant which is assumed to be of minimum phase and of an arbitrary relative degree. In the adaptive law the output error plays a dual role in the adjustment of the control parameter vector. In the absence of unmodeled dynamics and disturbances, the tracking error is shown to be zero when reference input is persistently exciting.