Highlights • Real-time control of cascading failures of large-scale bus transit networks. • An emergency route relinking approach and strategies. • Uncover relationship between preventive and emergency controls. • Insights into real-world emergency management in case study.

Abstract Cascading failures of large-scale bus transit networks (LBTNs) often lead to extreme rather than general network performance issues from a network dynamics perspective. Previous studies mostly concentrated on modeling the unfolding process of cascading failures, while little attention was given to controlling such failures, particularly under emergencies. This paper addresses an open problem: is it preferable to employ a preventive control before cascading failures, or to implement an emergency control for real-time cascading failures? Specifically, an emergency route relinking approach that dispatches internal resources (redundancy) without adding external resources (investment) is proposed to handle the operational disruptions caused by cascading failures. Remarkably, an intermediate process, namely non-emergency route relinking, is formulated to reconcile the different study frameworks between the two types of controls, premised on clarifying their differences and interdependencies. This formulation ensures the transition from preventive to emergency controls, as well as from traditional link addition to emergency route relinking. Furthermore, utilizing topological knowledge from network science and operational knowledge from transportation engineering, non-emergency and emergency route relinking strategies are designed to compare the control performance between the two types of controls and among emergency strategies. Case results indicate that (i) The emergency control is better than the preventive control, and operational knowledge is more effective than topological knowledge; (ii) It is insufficient to solely regulate real-time unfolding paths by evacuating failure loads to bypass specific areas, attention should also be given to planning redundancy to reduce the scale of failure loads and handle the adverse fluctuations in failure strength caused by the redundancy dispatching process. This work offers valuable insights into real-world emergency management and an underlying simulator with verified domain knowledge for intelligent control in mega-scale networks.