Traffic-State-Conditioned Topological Vulnerability Assessment of Urban Connected-Vehicle Networks
Abstract
Urban vehicle density changes the spatial organization of potential vehicle-to-vehicle (V2V) links, yet a denser topology need not respond to node loss in one direction across communication settings. We assess this relationship with processed AVI-derived trip demand from Xuancheng, reconstructed routes, CityFlow-simulated traffic states, and undirected distance-threshold potential V2V graphs. Eighteen distinct date/windows are paired within six exact-clock blocks and evaluated at four link radii. Routeable departures are strongly associated with simulated mean speed (Spearman $\rho $ = −0.936) and density/accumulation ( $\rho $ = 0.961), while higher-versus-lower demand changes graph size and connectivity. Structural vulnerability, measured by largest-component and normalized-efficiency area under adaptive batch removal curves, has configuration-dependent direction. Among eight tested strategies, global-efficiency centrality (GEC) is the strongest tested generalist, belonging to the winner set in 201 of 288 repeated cells, but it incurs substantial computation. Route and missing-demand perturbations alter fine strategy order. Two building-aware static link interpretations retain about 98% of baseline edges yet increase component counts and largely preserve the tested five-method rank landscape. These findings concern potential topology and structural stress response; they do not estimate packet delivery or radio performance.