An adaptive online monitoring approach for networked functional urban traffic flows

As rapid urbanization and population growth intensify, traffic disruptions in cities have become increasingly frequent. Modern sensing technologies provide the technical foundation for real-time monitoring of road traffic. However, urban road networks comprise numerous links and require the collection of massive amounts of data, while the computational resources available for data acquisition and analysis are often limited. Therefore, under constrained computing budgets, it is crucial to optimally select roads for sampling so as to improve the accuracy of real-time anomaly monitoring across urban networks. This work proposes an online monitoring method for traffic anomalies based on adaptive sampling algorithm. Traffic flow is modeled using a Network Functional Autoregressive Model (NFAR), which captures both temporal and spatial dependence in a unified framework. The Functional Principal Component Analysis (FPCA) is employed to extract data features. Based on the estimated parameters and the extracted features, monitoring statistic is constructed to inform sampling strategies. Numerical study shows that the proposed monitoring approach can fit the time series networked data well and performs well in anomaly detection.