Graduation Year

2022

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Civil and Environmental Engineering

Major Professor

Yu Zhang, Ph.D.

Committee Member

Fred L. Mannering, Ph.D.

Committee Member

Peng Chen, Ph.D.

Committee Member

Susana K. Lai-Yuen, Ph.D.

Committee Member

Ying Chen, Ph.D.

Keywords

Shared Mobility, Travel Behavior, Transportation Equity, Demand Forecasting, Data Driven Modeling

Abstract

With the adoption of advanced information technologies, the transportation system keep innovating. It now offers more travel options, becomes smarter, but meanwhile gets more complicated and imposes more challenges to researchers and practitioners. For example, demand responsive public transit operates according to user demand instead of traditional timetable-based and fixed route services, which brings a lot of convenience to users. These emerging services also reshapes the way people travel. With the popularity of ride sharing services such as Uber and Lyft, people are more accessible to vehicle use without an ownership. Given the increasing popularity of emerging services, it is important to understand it and manage with the aid of the latest algorithms.

The objective of this study is to explore different advanced data-driven methodologies that can be applied to identify the transportation ‘pain spots’ in a city, e.g., areas lack of transportation infrastructure or services, understand emerging transportation systems and improve its services and equity performance. Specifically, this study addresses the following three topics:

First, from transportation supply perspective, a concept called “multi-modal deserts” is proposed, which refer to areas with limited transportation infrastructure and mobility services that constrain people from accessing and taking different transportation modes. The transportation infrastructure and mobility services considered in this study include road network factors, public transit, shared micromobility and household vehicle ownership. To identify multi-model deserts, multivariate outlier detection method is applied to determine if some areas are significantly deviated from other areas considering the quality of aforementioned transportation infrastructure and mobility services. Downtown Tampa, Florida is selected as an empirical study area to demonstrate the proposed method. In total, 11 multi-modal deserts are identified among 182 Census Block Groups, and most of them are at edge of study area. The 11 multi-modal deserts are grouped according to social-demographic features of these areas and the characteristics of transportation supplies are analyzed. This study also tests the case if shared micromobility services were not deployed in the case study city. Results show shared micromobility can be an effective approach to improve mobility of areas with high poverty ratios and thus transportation equity. The results of the study help local authorities understand the mobility gaps in the city and support their decision on allocating resources for improving equal access to urban mobility for all citizens.

Second, we further study shared micromobility and explore factors influencing the usage of shared e-scooters, one of fast growing and most popular shared micromobility services in last several years. We also investigate the factors influencing users’ decisions of taking shared e-scooters to replace automobile mode, such as driving or taking taxi/TNCs. As revealed in the previous study, shared e-scooter program not only provides first-mile-last-mile solution, but also help improve transportation equity. Limited understanding of shared e–scooters restrain policymakers from developing more effective regulations and promoting this sustainable transportation mode. In our study, survey data are collected from shared e-scooter users, and random parameter models are applied to explore the factors influencing e-scooter sharing usage and mode substitution. Factors considered in models include sociodemographic information, user behaviors, trip purposes, and health indicators. Model results identify several factors that significantly influence shared e-scooter usage including user gender, helmet use, exposure to shared e-scooters, ownership of an e-scooter, riding locations, opinions on speed limits, and trip purposes. Results of auto substitution model suggest that shared e-scooters potentially are competing with TNC/taxi, lower cost, and social/entertainment trip purpose. User households with multiple vehicles contribute to private vehicle substitution. Research outcomes suggest that shared e-scooters could play a significant role in urban transportation sustainability.

Third, we turn our attention to shared mobility, Taxi and TNCs. In the first study in this dissertation research, unfortunately, the case study does not have Taxi and TNC data public accessible. Although it is a caveat of the case study, the concept of multimodal desert proposed, and the multivariate outlier detection method applied can be used for other cities with more comprehensive and complete data sources. Meanwhile, transportation demand forecasting is essential in urban transportation system. An accurate short-term forecast of passenger demand can help operators of Taxi and TNCs better allocate resources to achieve supply demand balance in real time. It improves efficiency of the services and reduce wait time of users. For socially disadvantaged areas with poor transportation services, it can be of great use. We propose a multi-task learning model that jointly predicts two-related transportation modes taxi and TNCs. In the model, the demand information between taxi and TNCs are shared through a gating mechanism. The gates selectively share useful information among two forecasting tasks to avoid negative information transfer. The technique is tested with Taxi and TNC demand in Manhattan, New York. The prediction accuracy of single-task learning, and multi-task learning models are compared, and the results show that the multi-task learning approach outperforms single-task learning and other benchmark models.

To summarize, this study applies advanced data-driven approaches to identify areas with inadequate transportation services, understand emerging shared e-scooters and forecast short-term demand of shared mobility. The outcomes provide insights to policy makers, operators, and urban planners.

Download

Share

COinS