A Two‐Dimensional RMPC Approach for Heavy‐Haul Trains Considering Model Uncertainty

ABSTRACT

To address the critical challenges of uncertainties and multi‐objective optimisation in heavy‐haul train control, this paper proposes a novel two‐dimensional robust model predictive control (2‐D RMPC) framework that uniquely integrates iterative learning and robust control principles. By comparing actual train measurement data with calculated results from the existing empirical model, the bounds of unknown disturbances are analysed and a train motion equation incorporating uncertainties is formulated. On the iteration axis, an online parameter identification algorithm is developed that exploits the repetitive nature of train operation and accounts for all unobservable relative motions between adjacent vehicle units, then its convergence is rigorously proved. Through combining historical and real‐time data, this algorithm dynamically provides updated train model parameters to the RMPC controller. On the time axis, the RMPC actively suppresses unknown time‐varying disturbances while optimising speed tracking, energy efficiency, and operation smoothness with multiple objectives and constraints. Case studies demonstrate that the identification algorithm can achieve convergence at the iteration domain and provide more precise train model parameters for RMPC to improve control accuracy. Compared with RMPC and MPC, the proposed method has illuminated superior performance in both speed tracking and anti‐disturbance.