This paper deals with the robust control of an Autonomous Underwater Vehicle (AUV) subject to communication constraints. The aim is the design of a gain-scheduled varying sampling controller using non periodic measurements from ultrasonic sensors. The contribution of this work lies in a new model formulation of varying sampling discrete-time systems using the Linear Fractional Transformation (LFT) representation. This allows to design, in the LFT context, an Hinf controller whose matrices are scheduled w.r.t the varying control interval. Moreover the proposed design ensures a performance adaptation when the sampling interval varies. This methodology is developed for AUV and emphasizes important improvements compared with an Hinf discrete-time control with fixed sampling.