In the high-speed train control system, it is possible to realize the mutual real-time communication between trains and ground equipments, thus the real-time information about trains can be transmitted to the ground commanding center. Under this new operation paradigm, in order to improve its safety and efficiency, this paper proposes the generalized and hierarchical framework of model predictive control (MPC) for the railway system including macroscopic, mesoscopic and microscopic levels. Under this framework, this paper further elaborates the coordinated following control based on MPC among adjacent trains in order to guarantee proper safety distance in case of unexpected disturbances. The Levenberg-Marquardt optimization approach is utilized to engender the corresponding control commands. The simulation results demonstrate the efficiency and robustness of MPC with the prediction models of trains’ movement for the coordinated control among them.