Envisioning Shared Autonomous Vehicles (SAVs) for 374 Small and Medium-Sized Urban Areas in the United States : The Roles of Road Network and Travel Demand

This study envisions a future of operating shared autonomous vehicles (SAVs) to provide mobility services in 374 small and medium-sized urban areas (population < 375,000) in the United States. Specifically, this study 1) generates large-scale synthetic commuting trip data for 374 urban areas, 2) employs an agent-based modeling framework to simulate the operations of SAV fleets serving commuters in these areas, and 3) develops models to explain the variations of fleet performances across study areas. The main objectives of this study are to compare the performances of SAVs across urban areas and to identify the correlates of SAV fleet performances related to regional road network and travel demand characteristics. This study generated 30 million synthetic commuting trips for 374 small and medium-sized urban areas based on the Census’ latest Longitudinal Employer-Household Dynamics (LEHD) data and the National Household Travel Survey (NHTS) data. Based on the agent-based modeling results, SAV fleet performances across 374 urban areas were analyzed. Key performance measures include the average number of trips per SAV vehicle for fleet efficiency and the percentage of empty vehicle miles travel (VMT) for extra operating cost. Results showed significant spatial variations of SAV fleet performances across areas. Urban areas in the Midwest and/or Great Plains appeared to associate with a greater SAV fleet efficiency than those in the Southeast, especially the areas along the Appalachia Mountains. The results highlighted urban areas that either outperform or underperform other areas. For example, Great Falls MT is associated with the highest fleet efficiency and Fort Walton Beach-Navarre-Wright FL has the lowest percentage of empty VMT. Through statistical modeling, this study identified significant correlates of SAV fleet performances which are significantly related to factors of road network properties and travel demand characteristics. A larger, denser and more connected network is associated with a greater fleet efficiency and reduced operating cost per unit service while an irregular network (e.g., more dead ends) or highly vulnerable network (e.g., bridges and bottleneck roads) may lead to decreased fleet efficiency and increased operating cost. Areas with denser trips or more long trips are associated with lower fleet efficiency and decreased operating cost. More modeling results are discussed in the paper to provide valuable information for localizing strategies for operating SAVs to consider local road network and travel patterns. This study complements the literature of shared mobility and automation research currently centered on major urban areas. The research findings are intended to serve as the basis for future discussions regarding opportunities and challenges of deploying emerging mobility services in small and medium-sized areas. The large-scale synthetic commuting data generated in this study are valuable for researchers and agencies to develop advanced travel and mobility models and to further explore the potentials of deploying emerging mobility systems in these areas