Shared Mobility - Operations and Economics

Abstract

In the last decade, ubiquity of the internet and proliferation of smart personal devices have given rise to businesses that are built on the foundation of the sharing economy. The mobility market has implemented the sharing economy model in many forms, including but not limited to, carsharing, ride-sourcing, carpooling, taxi-sharing, ridesharing, bikesharing, and scooter sharing. Among these shared-use mobility services, ridesharing services, such as peer-to-peer (P2P) ridesharing and ride-pooling systems, are based on sharing both the vehicle and the ride between users, offering several individual and societal benefits. Despite these benefits, there are a number of operational and economic challenges that hinder the adoption of various forms of ridesharing services in practice. This dissertation attempts to address these challenges by investigating these systems from two different, but related, perspectives.

The successful operation of ridesharing services in practice requires solving large-scale ride-matching problems in short periods of time. However, the high computational complexity and inherent supply and demand uncertainty present in these problems immensely undermines their real-time application. In the first part of this dissertation, we develop techniques that provide high-quality, although not necessarily optimal, system-level solutions that can be applied in real time. More precisely, we propose a distributed optimization technique based on graph partitioning to facilitate the implementation of dynamic P2P ridesharing systems in densely populated metropolitan areas. Additionally, we combine the proposed partitioning algorithm with a new local search algorithm to design a proactive framework that exploits historical demand data to optimize dynamic dispatching of a fleet of vehicles that serve on-demand ride requests. The main purpose of these methods is to maximize the social welfare of the corresponding ridesharing services.

Despite the necessity of developing real-time algorithmic tools for operation of ridesharing services, solely maximizing the system-level social welfare cannot result in increasing the penetration of shared mobility services. This fact motivated the second stream of research in this dissertation, which revolves around proposing models that take economic aspects of ridesharing systems into account. To this end, the second part of this dissertation studies the impact of subsidy allocation on achieving and maintaining a critical mass of users in P2P ridesharing systems under different assumptions. First, we consider a community-based ridesharing system with ride-back guarantee, and propose a traveler incentive program that allocates subsidies to a carefully selected set of commuters to change their travel behavior, and thereby, increase the likelihood of finding more compatible and profitable matches. We further introduce an approximate algorithm to solve large-scale instances of this problem efficiently. In a subsequent study for a cooperative ridesharing market with role flexibility, we show that there may be no stable outcome (a collusion-free pricing and allocation scheme). Hence, we introduced a mathematical formulation that yields a stable outcome by allocating the minimum amount of external subsidy. Finally, we propose a truthful subsidy scheme to determine matching, scheduling, and subsidy allocation in a P2P ridesharing market with incomplete information and a budget constraint on payment deficit. The proposed mechanism is shown to guarantee important economic properties such as dominant-strategy incentive compatibility, individual rationality, budget-balance, and computational efficiency.

Although the majority of the work in this dissertation focuses on ridesharing services, the presented methodologies can be easily generalized to tackle related issues in other types of shared-use mobility services.