System-level impact and behavior of coordinated vehicle fleets in transportation networks

Over the past decade, travelers have increasingly utilized technology to aid their use of transportation systems, from real-time navigation apps like Google Maps, to ride-sourcing apps like Uber and Lyft, to bike and scootershare apps like Lime and Spin. These applications of technology to mobility systems have fundamentally changed the way in which people travel by offering them the ability to navigate the physical world by first navigating a virtual representation of it. The layers of virtual and physical transportation infrastructure complicates transportation planning efforts and underscores the need for methodologies to understand and leverage the relationship between the virtual and physical transportation infrastructures. This dissertation offers concrete methodologies to understand the behavior and impact of technologically-enabled mobility services on transportation infrastructure along three directions: 1) to illuminate the coordinating signals in virtual mobility services, 2) to anticipate benefits and harms of coordinating signals in general physical transportation networks, and 3) to recover coordinating signals from observed behavior on the physical network. These methodologies together expand the toolbox of the transportation planner to not only accommodate but leverage technological change in realizing the transportation system of the future. In the first part of this thesis, we demonstrate that spatial-temporal imbalances in supply and demand on ride sourcing platforms can be predicted from real-time urban data. This work highlights surge pricing as an imperfect market mechanism and also demonstrates that ride-sourcing supply and demand behavior in the aggregate can be recovered from data. In the second part of the thesis, we study how coordination within a virtual transportation infrastructure impacts travelers on the physical road network. In the best case, this coordination induces the system optimal network utilization, but in other cases may reduce network efficiency. Finally, we study how we might infer the coordinating signal within a virtual transportation network from observed behavior of its users on the physical network, even as the environment is changing. To this end we apply a novel statistical test to driver behavior during the COVID-19 lockdown in Pittsburgh. This research may also be applied to assess the validity of network equilibrium models based on a particular notion of travel cost to describe the longrun behavior of a network from data collected on that network over time. Taken together this thesis outlines a methodology for transportation planners to recover coordinating signals from data and leverage them in longterm transportation plans to improve transportation systems.