Iterative Learning Model Predictive Control for Robust Rescheduling of Intercity Express Trains

In this paper, we aim to achieve robust and efficient train rescheduling of intercity express railway lines considering periodic train timetables and passenger uncertainties. Considering intercity express railways’ features of longer section lengths, multiple trains running in close succession within long sections and the varying passenger demands, we improve the multi-train state space model of intercity express railway operation, in which the changes and uncertainties of passenger flows are modeled as system parameter uncertainties and the primary train delays caused by temporary speed restriction extreme weather, and signal failure, etc. are modeled as external interference. Then, a real-time train rescheduling controller is developed that combines iterative learning and model predictive control to enhance its resistance against both the model uncertainties (i.e., varying passenger flows) and the external interference (i.e., the primary delays). The objective function of the rescheduling controller is to recover from delays while preventing the control force amplitude from becoming excessively large. Each period of the periodic timetable is modeled as a batch, and a batch-based state space error predictive model is developed to simultaneously recover the nominal timetable and minimize the control force amplitude. It is proven that the error norm will eventually converge to a bounded value as the number of iterations increases. The performance of the proposed method is evaluated through simulations based on the Beijing–Tianjin intercity express rail line.