Supply-Demand Equilibrium of Private and Shared Mobility in a Mixed Autonomous/Human Driving Environment

This dissertation research presents a methodological investigation of the transportation system performance with coexistence of privately owned human-driven and autonomous vehicles (HVs and AVs) and shared AVs (SAVs) that provide ridesharing services. Built upon the network optimization and game theory concepts, the proposed models holistically account for the complex interplays among three players which contribute to the system performance. On the supply side, car manufacturers will sell (S)AVs at possibly higher prices than that of HVs due to the new technologies used in (S)AVs. In addition, Uber-like transportation network companies (TNCs) will affect the level of ridesharing service directly through the SAV supply attributes (such as fare and fleet size/allocation/relocation) and indirectly via the efficiency of the matching technology used for establishing SAV-traveler contacts which manifests through service waiting times. On the demand side, travelers’ adoption of HVs and (S)AVs depends on their perceptions of the associated costs, which are not only affected by the endogenous strategies of car manufacturers and TNCs, but also by the congestion effects of occupied and unoccupied vehicle flows across road network. The supply of and demand for (S)AVs further interact with transportation infrastructure through changes in road network capacity and parking demand. The former is caused by shorter car following headways of automated/connected driving technologies used in (S)AVs compared to HVs. The latter is attributed to the self-parking capability of AVs in distant/cheaper areas and the reduced number of private cars in the presence of ridesharing. Such supply-demand-infrastructure interactions will be more complicated in mixed traffic of (S)AVs and HVs, which is expected to dominate at least for a few decades until human driving becomes obsolete. Besides the novelties in the presented models, this research also contributes to the field by developing an efficient disjunctive programming based global optimization algorithm to cope with the non-convex nature of the network optimization problems.