Abstract

This paper addresses the problem of modeling stochastic dynamic transportation networks, where travel times are random variables with time-varying distributions and spatio-temporal dependencies. It presents a taxonomy of modeling approaches used in the literature and defines extended taxonomic categories to better capture the characteristics of transportation networks. A two-part approach is presented to characterize transportation networks spatially and temporally based on the network structure and observed travel time data. The first part implements a community detection approach from network science modified for the application to transportation networks. The community detection approach is further adapted to use dynamic time warping distance to measure the dissimilarity between time series of link travel times. The second part implements a change point detection approach that identifies points in time when there are changes in the underlying link travel time distributions generating the observed time series. The proposed approach is tested via sensitivity analysis using the large-scale Chicago network and its corresponding travel time data. Sensitivity analysis with varying model parameters demonstrates the robustness of the network characterization and helps determine the most suitable parameter values. The sensitivity analysis with respect to varying weather conditions and demand patterns in the network demonstrates that the spatial and temporal characteristics are robust with respect to changing exogenous and endogenous conditions.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call