Network-wide Time-dependent Link Travel Time Distributions with Temporal and Spatial Correlations

Most studies in the literature on stochastic networks are based on hypothetical link travel time distributions. Absence of actual observations is one of the main obstacles for studying spatial and temporal link travel time correlations in general networks. To address this gap, a simulation- based dynamic traffic simulator is used to simulate different scenarios based on the real world observations such as weather conditions, freeway daily throughput, and incidents. In addition to link travel time distributions, different correlation structures are also explored in this study. The methodology is successfully applied on the large-scale Chicago network to estimate network-wide link travel time distributions and correlations. Through the numerical results, the existence and magnitude of the spatial and temporal link travel time correlations is evaluated. Furthermore, the static and dynamic link travel time distributions are compared and it is shown that assuming static link travel time distributions may result in biased conclusions.