In this paper, a robust control design using the neural network (NN) and dynamic surface control (DSC) method is proposed for path following of underactuated surface vessels with input saturation. First, the Serret-Frenet (SF) frame and the line of sight (LOS) method are used to transform the kinematics of ship motion into "lower-triangle" form; then the backstepping method can be applied directly. Second, the problem of complexity explosion inherent with traditional backstepping is avoided by employing DSC technique, and the computational burden is reduced. Third, the NN is included to cope with the uncertainties of the model. Fourth, an auxiliary design system is used to deal with input saturation of rudder, which can help reduce the abrasion of the rudder machine. The uniform boundedness of all the closed loop signals is guaranteed via Lyapunov analysis. The simulation results with comparisons are presented to verify and illustrate the effectiveness of the proposed controller. (C) 2017 Elsevier B.V. All rights reserved.