Predictive Control based on Bayesian Optimization for Station Parking of Trains with Discrete Gears

A predictive control method based on Bayesian optimization for station parking of trains with discrete gears is proposed in this paper. The nonlinear and complicated but essential elements of the train model are considered to get a more accurate prediction than the simplified models used in most of the existing control methods. Firstly, the train station parking problem, which takes parking accuracy, punctuality and riding comfort into account at the same time, is reformulated as an optimization problem to find the optimal speed switch points given a gear sequence based on the current train state. Then Bayesian optimization is utilized to solve the optimization problem. Lastly, the predictive control method is proposed, which can park the train within the required stopping error by a few gear shifts and improve the riding comfort. Moreover, the method is flexible under different station parking time constraints. The simulations are performed to show the effectiveness of the proposed method.