Critical multi-link disruption identification for public transport networks: A multi-objective optimization framework

Public transportation networks are vulnerable to uncertainty, which manifests in various forms, disrupting their operations and leading to delays and passenger dissatisfaction. Strategic-level planning of public transport networks should include the identification of critical disruption scenarios that may result in the loss of network functionality and increased travel times for users. Existing studies on transit network vulnerability have focused on identifying isolated critical links, overlooking simultaneous failures and their impacts. In this context, this study presents a multi-objective algorithm based on Adaptive Variable Neighborhood Search (MO-AVNS) to identify critical disruption scenarios affecting transit network serviceability, using a transit assignment model to capture passenger reactions to these. A set of critical combinations is generated, reflecting transit network link failures that maximize unsatisfied demand and additional travel time, capturing both users without viable travel options and passengers whose shortest-paths are disrupted. Results on a test network are presented for scenarios featuring up to five simultaneous link failures and compared to those based on centrality-based attacks. Empirical findings demonstrate that optimization-based attacks can identify scenarios that result in significant shares of disconnected passengers and high detour costs for the remaining passengers, that would be missed under single-objective approaches or centrality metrics.