Distributed Model Predictive Control for Automatic Train Regulation in Metro Lines

In high-frequency metro lines, unexpected disturbances often occur to cause train delays, which will influence the application of pre-optimized timetables. Real-time train regulation adjusting running and dwell times is an effective method to suppress train delays. This paper investigates a real-time metro train regulation method combining the distributed model predictive control (DMPC) algorithm, which is motivated by the technology of vehicle-based train control system. In the vehicle-based system, each train is self-organized with the capability of communication and computation. Firstly, an optimal control model is established for the metro train regulation process considering the train traffic dynamics, system constraints and objectives. Then, a distributed control framework is proposed to solve the optimal control model. In addition, DMPC algorithms with no information transmission, single-direction and bi-direction information transmission are designed, aiming to analyze the influence of the communication modes on the DMPC algorithms. Finally, the proposed algorithms were subjected to numerical validation to assess their efficacy. The results show that the DMPC algorithm with bi-direction information transmission achieved significant performances in reducing computation costs and suppressing train delays.