Author ORCID Identifier

0000-0003-1286-0318

Document Type

Dissertation

Date of Award

8-31-2023

Degree Name

Doctor of Philosophy in Transportation - (Ph.D.)

Department

Civil and Environmental Engineering

First Advisor

Branislav Dimitrijevic

Second Advisor

Lazar Spasovic

Third Advisor

I-Jy Steven Chien

Fourth Advisor

Joyoung Lee

Fifth Advisor

Athanassios K. Bladikas

Abstract

The advent of autonomous urban mobility is expected to effect changes in land use patterns and urban development. A new framework for integrated land use and transportation system models that incorporate Autonomous Vehicles (AVs) and Shared Autonomous Vehicles (SAVs) will be needed to investigate the development of transportation and land use plans and policies. This study aims to analyze the impacts of AVs and SAVs on land use changes by defming scenarios that consider different shares of these emerging transportation modes. Also, this study explores the corollary effects of various land use factors by applying them to a case study area in the State of New Jersey. The data uses in the analysis consists of household attributes and zone attributes, including household travel surveys, loaded highway networks, housing unit density, socio-economic factors, home values, school scores, and crime scores. Three methods are employed for land use models, including Multinomial Logistic Regression (MNL), Agent-Based Model (ABM), and k-means clustering method with Principal Component Analysis (PCA) applied to the MNL model. The models are implemented using a household survey dataset containing records of 7,896 households in the State of New Jersey. Using a static traffic assignment, the study applies a four-step travel demand model to find the travel times and travel costs between zones. The resulting impedance matrix is used to measure the zone accessibility, including residential and job accessibilities, which are then used in the land use models. Two scenarios are introduced to test the models: (a) scenario I, which considers only Conventional Vehicles (CVs), and (b) scenario II, which considers market penetration of 50% CVs, 25% AVs, and 25% SAVs. To test the model performance, scenario I is compared to the actual conditions showing the surveyed households' locations. The results reveal that the k-means clustering method with PCA applied on MNL outperforms two other methods since homogeneity and addressing the multicollinearity of data can improve model prediction accuracy, which is 0.86. Additionally, according to the MNL results in scenario I, which only considers CVs, households with high incomes (above $150,000 annually) have a negative inclination to relocate, which may indicate that they currently reside in desirable areas. It is found that in the model, the family structure impacts the chance of relocation; specifically, single workers without children are more likely to desire to move than two or more employees in households with children. As expected, the results indicate that households prefer locations with better transit, less crime, and higher school scores, and those with fewer vehicles have a greater tolerance for relocating. Furthermore, the results from scenario I are compared to scenario II in terms of median and mean distances to the central business district (CBD). The findings show that, in scenario II, the mean and median distances to CBD are increased in all applied methods. This occurs as a result of some households moving away from CBD in search of higher utilities, like more reasonably priced housing. Moreover, the sensitivity analysis of location decisions is conducted using the best-performing land use model (k-means clustering method with PCA applied on MNL), considering the variation in AV and SAV market penetration and the variation in the perceived value of travel time when using the two autonomous vehicle modes, respectively. According to the findings, the increase in the proportion of SAVs will cause more households to move further from the CBD.

Furthermore, with market penetrations of 50% AVs and 50% SAVs, households tend to move to suburbs, with the mean distance of household locations from the CBD exceeding 36 miles. Also, the median and mean distances to the CBD increase with the reduced perceived value of travel time when using AV or SAV. This occurs due to the increased opportunity for households to utilize AVs and SAVs for longer commuting trips while residing in more attractive zones farther from the CBD. The demonstrated models can inform the development of a policy framework considering different aspects of integrated land use and AVs and SAVs implementation in transportation networks. The proposed modeling framework and the resulting plans and policies will help ensure better coordination and collaboration among relevant stakeholders in advancing regional mobility solutions, addressing infrastructure needs, and improving safety and security in anticipation of the advent of AV and SAV.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.