This paper investigates the formation control problem of underactuated unmanned surface vehicles with unknown dynamics and ocean disturbances. To tackle the leader-follower configuration, an improved fixed-time velocity controller is proposed for the virtual vehicle to guide the followers efficiently. For the tracking control of the follower vehicles, a concise controller is designed by borrowing on backstepping and nonsingular sliding mode control techniques. A novel adaptive filter is developed to address the so called “explosion of complexity” problem in the traditional backstepping framework, which could simplify the control structure and facilitate the implementation of the proposed controller in practical engineering. Furthermore, the radial basis function neural network is employed to identify the unknown dynamics of the vehicle model. A bounded-feedback adaptive law is developed to estimate the upper bound of the ocean disturbances. That could greatly enhance the robustness and accuracy of the controller. Finally, a rigorous proof has been given to guarantee the practical fixed-time stability of the closed-loop system. Two examples are provided to demonstrate the effectiveness and superiority of the theoretical results.