Distributed Model Predictive Control for Real-Time Automatic Train Regulation of Metro Networks With Transfer Connections

To improve the operation and service efficiency of automatic train supervision systems in metro networks, this paper investigates the distributed train regulation control problem for metro networks under disturbances. Train-to-Train (T2T) communication technology, a key component of next-generation Communication-Based Train Control (CBTC) systems, has fundamentally shifted industry and academic focus in the field toward decentralized, adaptive, and infrastructure-light solutions. Leveraging T2T communication, we propose a distributed control system architecture to alleviate the computational burden of central traffic controllers. More precisely, each train is equipped with an onboard controller that makes decisions based on the state of the network, and a novel real-time train regulation strategy is developed, taking into account the metro’s operational and service constraints. The proposed approach aims to minimize traffic delays and optimally coordinate connections between trains of different lines. The goal is also to reduce the waiting time deviation for transferring passengers. Unlike existing studies that primarily focus on centralized approaches and single open lines, we model both one-direction loop lines and two-track open lines as loop lines and then propose a Distributed Model Predictive Control (DMPC) approach to address the regulation challenges of loop line configurations. Numerical simulations are performed on the Beijing metro network using real data, and the results are discussed to demonstrate the effectiveness and efficiency of the proposed automatic train regulating method.