Adaptive safety management of bidirectional crowd in metro stations considering robustness: From data-driven identification to prediction control

In the high-density bidirectional crowd environment of subway stations, dynamically adjusting diversion railings in the channel is a refined management strategy to improve traffic efficiency and safety. At present, railings usually adopt a fixed layout or are manually adjusted by staff based on experience, which makes it difficult to achieve adaptive management. To address this issue, this paper proposes a dynamic control framework for bidirectional crowds based on data-driven crowd dynamic system identification and model predictive control (MPC). The pedestrian dynamics theory is used to build a three-dimensional model of the bidirectional crowd under railings to generate a high-quality data set. Three types of data-driven linear and nonlinear identification models are constructed, and indicators such as FPE, NRMSE, MSE, FPE, AIC, and BIC are introduced to evaluate the accuracy of identification results. Based on the identification model, the MPC controller is designed with the railing position as the control input and the crowd density difference on both sides of the railing as the control target. The robust performance of the optimization strategy is ensured by setting the response limit of the control output. The traffic quality assessment model is developed to evaluate walking efficiency and safety. Simulation data shows the railing control strategy taking into account robustness significantly balances the traffic smoothness and safety and has a certain anti-interference ability. In addition, the MassMotion simulation system further demonstrates the ability of the proposed optimization strategy. This method provides a novel solution to the problem of high-density bidirectional traffic safety management, and provides a practical guide for station managers’ decision-making.