Abstract

Shared mobility systems have gained significant attention in the last few decades due, in large part, to the rise of the service-based sharing economy. In this thesis, we study the matching mechanism design of two-sided shared mobility systems which include two distinct groups of users. Typical examples of such systems include ride-hailing platforms like Uber, ride-pooling platforms like Lyft Line, and community ride-sharing platforms like Zimride. These two-sided shared mobility systems can be modeled as two-sided markets, which need to be designed to efficiently allocate resources from the supply side of the market to the demand side of the market. Given its two-sided nature, the resource allocation problem in a two-sided market is essentially a matching problem.

The matching problems in two-sided markets present themselves in decentralized and dynamic environments. In a decentralized environment, participants from both sides possess asymmetric information and strategic behaviors. They may behave strategically to advance their own benefits rather than the system-level performance. Participants may also have their private matching preferences, which they may be reluctant to share due to privacy and ethical concerns. In addition, the dynamic nature of the shared mobility systems brings in contingencies to the matching problems in the forms of, for example, the uncertainty of customer demand and resource availability.

In this thesis, we propose matching mechanisms for shared mobility systems. Particularly, we address the challenges derived from the decentralized and dynamic environment of the two-sided shared mobility systems. The thesis is a compilation of four published or submitted journal papers. In these papers, we propose four matching mechanisms tackling various aspects of the matching mechanism design. We first present a price-based iterative double auction for dealing with asymmetric information between the two sides of the market and the strategic behaviors of self-interested agents. For settings where prices are predetermined by the market or cannot be changed frequently due to regulatory reasons, we propose a voting-based matching mechanism design. The mechanism is a distributed implementation of the simulated annealing meta-heuristic, which does not rely on a pricing scheme and preserves user privacy. In addition to decentralized matching mechanisms, we also propose dynamic matching mechanisms. Specifically, we propose a dispatch framework that integrates batched matching with data-driven proactive guidance for a Uber-like ride-hailing system to deal with the uncertainty of riders’ demand. By considering both drivers’ ride acceptance uncertainty and strategic behaviors, we finally propose a pricing mechanism that computes personalized payments for drivers to improve drivers' average acceptance rate in a ride-hailing system.