Travel Time Reliability in Stochastic Dynamic Transportation Networks: Modeling, Path Finding and Routing

Filipovska, Monika

doi:https://doi.org/10.21985/n2-561e-p641

Work

Travel Time Reliability in Stochastic Dynamic Transportation Networks: Modeling, Path Finding and Routing

Public

Download PDF

Download All Files (.zip)

Travel time is a key aspect of capturing and evaluating the operational performance and service quality of transportation systems, and travel time improvement is a common objective for travelers, service providers, transportation practitioners and agencies. However, the reliability of travel times, including the probability of unexpected delays, is an important factor for travelers’ satisfaction with their transportation network experience, shown to affect their travel choices. Travel time reliability can also play a significant role in the decision making of goods shippers and logistic firms, where variability can have a direct economic impact. Modeling travel time variability is key to quantifying the inherent uncertainty in the knowledge of future events in a transportation network. It provides a more comprehensive representation of the state of the network and allows for making decisions that account for uncertainty and are robust to potential travel time variability. This dissertation is concerned with modeling, optimization, and analysis problems in stochastic dynamic transportation networks, where link travel times are modeled as random variables with time-varying distributions. Motivated by the need for data-driven and application-oriented modeling and optimization approaches for transportation network analysis that consider travel time reliability, the overarching goal of this thesis is to define, model and present solution approaches for key problems in stochastic dynamic transportation networks. The key objectives of this thesis include (1) modeling the temporal and spatial dependencies in stochastic transportation networks revealed in observed travel time data, (2) devising solution approaches for the estimation of path travel time distributions considering those spatio-temporal dependencies, (3) defining and solving path finding problems for reliable least-time routing, (4) incorporating en-route information in routing problems and analyzing its impact on travelers’ decision making. To meet the first objective, this thesis presents a comprehensive methodology for modeling the temporal and spatial aspects of stochastic transportation networks, as well as their intersection: the temporal variation of spatial characteristics and vice versa. To address the second objective, this thesis presents, tests, and evaluates a number of distribution estimation approaches that consider the network’s spatio-temporal characteristics. The third objective is at the center of three problem classes of concern in this dissertation: (1) a priori reliable least-time paths, (2) trajectory-adaptive reliable least-time strategies, and (3) information-adaptive reliable least-time routing. The fourth and final objective is concerned with a key characteristic of stochastic network models, namely that knowledge of future network states can be adjusted based on information of past and current states. This objective is met via the latter two of these problem classes that consider traveler decision making based on in-vehicle trajectory data availability and connected vehicle information access in a connected environment. In addressing these core objectives, this dissertation achieves the larger goal of presenting a comprehensive conceptual and methodological framework for modeling, estimation, and optimization in stochastic dynamic networks. The modeling and estimation methods from the first two objectives are key tools in addressing the problems in the following two objectives. Furthermore, the solution approaches for the three routing problems contain a shared component that allows for their solution to be initiated in a single shared procedure. The problem classes and solution approaches that this thesis is concerned with have several important application areas. The stochastic transportation networks characterization and estimation of path travel time distributions can be applied for performance measurement and monitoring of transportation policies, projects and applications concerned with the reliability performance of transportation systems. Reliable path finding problems have a host of relevant applications, such as reliability-based vehicle routing of freight or mobility service providers, or applications for emerging transportation technologies and services such as electric vehicles, autonomous vehicles, ride-sourcing companies, etc. Real time data access and the increased use of navigation services also call for making reliability-based decision-making adaptive to information.

Creator