Integrating Shared Autonomous Vehicle Fleet Services in Overall Urban Mobility: Dynamic Network Modeling Perspective

Abstract

Like many other domains, transportation is undergoing deep and significant transformation, seeking to fulfill the promise of connected mobility for people and goods, while limiting its carbon footprint. The advent of autonomous vehicles has the potential to change the economics ownership and use of private automobiles, likely accelerating trends towards greater use of app-based ride hailing and/or sharing by private TNCs (Transportation Network Companies). Several potential business models with varying degrees of ride sharing and public vs. private involvement in the delivery of mobility as a service (MaaS) are presented. Algorithms for shared autonomous fleet management are discussed and illustrated on a small case application. These are then integrated in an intermodal network modeling framework, applied to the Chicago region to evaluate the impact of new services on mobility and sustainability. By reinventing themselves as mobility agencies, public transit companies can leverage these developments to focus resources on providing high-quality services along high-density lines, resulting in significant improvement in overall urban and regional mobility. This paper focuses on an agent-based microsimulation of a transit urban network system with shared-ride autonomous vehicles (SAV) as first-mile feeders to assess SAV demand and its impact on transit demand. We introduce an integrated mode choice and dynamic traveler assignment-simulation modeling framework that explicitly models a transit network and SAV fleet system. We employ a bi-level and iterative solution approach due to the dependency of mode shares on each modeÕs performance, and the dependency of the modeÕs performance on modal flows. In the iterative modeling framework, the upper level assigns travelers to one of five modes: car, park-and-ride, transit, SAV, or transit with SAV feeder. The lower level, both (1) iteratively determines minimum cost transit hyperpaths, assigns travelers to hyperpaths, and simulates their experiences, and (2) simulates an SAV fleet providing service to travelers. Time-dependent network performance metrics are fed to the mode choice model, which reassigns travelers to modes to have their experiences simulated in the next iteration. This process repeats until the mode choice probabilities converge. This integrated modeling framework, which endogenously determines traveler mode choice as well as transit and SAV system performance, provides transportation planners and modelers a powerful tool to test various scenarios related to AV-enabled mobility services.