In this paper, based on a proposed MIMO hybrid dynamical system of underwater vehicle and its actuators, a robust composite adaptive fuzzy controller is presented. The proposed method employs voltage control effort, which is more efficient than the torque control strategy. Also it is very simple, efficient and robust. Based on adaptive fuzzy method and the prediction error between the system states and the serial–parallel estimation model, a composite adaptive fuzzy law that uses the modeling error as input is constructed to adaptively compensate the unknown uncertainties and disturbances of the system. In addition, the proposed scheme is able to estimate the lumped uncertainties with large amplitude and high frequency. Stability of the proposed method is shown based on Lyapunov approach. The proposed control scheme is not limited only to control of the underwater vehicles, but can be applied for a class of nonlinear MIMO systems with square and non-square control gains.