In this article, an efficient hierarchical control framework is proposed to address the cooperation problems (e.g., consensus tracking, formation tracking, and time-varying formation tracking) for the networked marine surface vehicles in the presence of external disturbances, actuator faults and failures.Based on this framework, several learning-based hierarchical control algorithms are developed, involving an iterative learning-based estimator and a local observer-based finite-time controller. The estimator is designed to achieve sufficiently precise estimation of the leader states through enough iterations, while the observer-based finite-time controller is used to observe and compensate the dynamic uncertainties as well as stabilize the error states in a finite time. By using the theories of Hurwitz, Schur, and Lyapunov stability, the sufficient conditions for guaranteeing the convergence of these learning-based hierarchical control algorithms are derived. Finally, numerical simulations are performed on the Cyber-Ships II to verify the effectiveness of the presented algorithms.