Design and optimization of shared mobility systems

Abstract

Rapid urbanization poses significant challenges to urban mobility. Shared mobility services, defined as the collective utilization of transportation resources, have emerged as a promising solution to these challenges. However, their implementation faces complex decision-making issues including vehicle routing, order assignment, pricing scheme design, etc., particularly under dynamic or uncertain conditions. Addressing these issues is crucial for fostering an efficient and sustainable mobility ecosystem. This thesis addresses three key decision-making problems within shared mobility services: dynamic vehicle dispatching for shared-and-autonomous-mobility (SAM) services incorporating ride-pooling, compensation scheme design for integrative shared mobility (ISM) services under stochastic demand, and public transit line planning (PTLP) with bike-sharing integration. The first research problem investigates dynamic vehicle dispatching for SAM services with ride-pooling options. An algorithmic framework based on a rolling horizon approach is proposed, continually updating vehicle dispatch plans based on real-time demand information by solving a series of static subproblems. Each static subproblem is formulated as a mixed-integer programming (MIP) model and solved by a customized hybrid algorithm, named ARA-LNS, which integrates an adaptive request assignment (ARA) into a large neighborhood search (LNS) heuristic framework to efficiently optimize the request assignment and vehicle routing plans. The second research problem explores the optimal compensation scheme design for ISM services that simultaneously provide both passenger ride and parcel delivery services using an on-demand shared vehicle fleet. To address the extra ride duration (ERD) caused by additional stops, the service operator compensates passengers, whose tolerance for ERD depends on the compensation amount. The problem is formulated as a two-stage stochastic programming model considering passengers' nonlinear acceptable ERD (AERD) profile and stochastic demands and solved by a sample average approximation method. A customized ALNS-CSA algorithm that combines an adaptive large neighborhood search (ALNS) heuristic and an efficient compensation scheme adjustment (CSA) method is developed to iteratively determine the optimal demand serving, passenger compensation, and vehicle routing (DPV) solution and improve the compensation scheme accordingly while respecting the AERD constraints. The third research problem focuses on the optimal design of the public transit line with integrated bike-sharing services to determine the optimal bus stop location and service frequency by minimizing total system costs, including both user and operator expenses. A simulation-based optimization modeling framework powered by a multi-agent-based simulation (MABS) system is developed to capture disaggregate behaviors and interactions of various entities in the bus operation system, especially incorporating the bike-sharing complementary feeder mode services. A surrogate-based optimization (SBO) solution method is introduced to solve the black-box simulation-based PTLP problem by efficiently approximating the mapping relationship between bus transit planning decision inputs and expected system cost output. This method allows us to identify high-quality stop location and service frequency solutions within a few objective function simulation evaluations. The efficacy of all the proposed models and solution methods for the three research problems is evaluated through extensive numerical experiments. Impact analyses of potentially influential factors are also conducted to derive managerial insights to guide the practical management and operations of shared mobility services.