Impacts of Autonomous Vehicles on Urban Land Use 

自動駕駛對城市土地利用的影響

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date1 Sep 2022

Abstract

Autonomous vehicles (AVs) have attracted much attention worldwide in practical applications and academic research. There is no doubt that AVs will bring dramatic changes to our transportation, lives, and society. And from horse-drawn carriages to automobiles to rail transit, every transportation change brings enormous changes to cities, including urban size and urban form. Exploring how AVs will affect urban land use is an important topic. At the same time, proposing strategies to cope with the changes is also a necessary measure to meet the arrival of the autonomous era.

Most scholars have applied travel models to simulate the potential impacts of AVs from a micro perspective, focusing on traffic volume, travel time, travel mode, etc. On the other hand, others have used questionnaires and interviews to measure the acceptance of AVs and residential choices qualitatively. In terms of urban form and land use, a few scholars have modelled changes in accessibility through scenario analysis. These studies lack a macroscopic view of the city as a whole, and thus the proposed response strategies are one-sided and partial. Furthermore, most of these studies take developed cities in North America and Europe as examples, but AVs may have a more significant impact on developing countries. Therefore, this study contributes to the impacts of AVs on land use from a macroscopic perspective and fills the gap in developing country studies by using Shenzhen as an example.

The land use and transport interaction (LUTI) model is a standard method to study transportation and land use problems, and system dynamics (SD) can precisely cope with the unsynchronized characteristics of the transportation system and the land use system in the LUTI model. This study uses SD to construct a LUTI model from qualitative and quantitative perspectives, incorporating urban systems such as population, economy, and employment. This SD-LUTI model lays the foundation for the following study.

With the rise of the sharing economy, sharing and carpooling are gradually becoming strong competitors to private cars, even more so in AVs. Through the established SD-LUTI model, four different scenarios are used to explore the different impacts of private AVs (PAV), shared AVs (SAV) and ride-sharing AVs (RAV). Based on some research results on travel time and travel mode, it is found that PAV induces more trips and increases congestion. In contrast, SAV and RAV are effective ways to mitigate this negative effect. What is more important is that AVs increase land demand, especially under shared transportation, which may induce urban sprawl.

To explore the micro spatial distribution of the increased land demand, accessibility, expressed by a gravity model, is first simulated in the temporal dimension using SD and then calculated in the spatial dimension using urban network analysis (UNA) on ArcGIS. Similar to the results in the previous section, PAV decreases the accessibility in the starting phase, while SAV and RAV increase accessibility. In terms of spatial distribution, the accessibility improvement is much more apparent in the suburbs than in the urban centre.

Finally, based on the above findings, this study proposes some planning strategies for AVs applications in three aspects. From the transport perspective, the implementation of shared SAV could be promoted. From the land use perspective, TOD can be implemented in the suburbs and sparse and dense development in the urban areas. From the planning strategy point of view, resilient planning should be implemented in comprehensive planning and special planning.

This study enriches the research perspective of AVs, regulates future land use planning from a more macroscopic perspective, and makes up for the lack of quantitative studies in developing countries. However, the research model is overly simplified, and the scenario assumptions have significant uncertainties, which need to be verified by more empirical data in the future.