Synthesizing neighborhood preferences for automated vehicles

Abstract

Automated Vehicles (AVs) have gained substantial attention in recent years as the technology has matured. Researchers and policymakers envision that AV deployment will change transportation, development patterns, and other urban systems. Researchers have examined AVs and their potential impacts with two methods: (1) survey-based studies of AV preferences and (2) simulation-based estimation of secondary impacts of varied AV deployment strategies, such as Shared AVs (SAVs) and Privately-owned AVs (PAVs). While the preference survey literature can inform AV simulation studies, preference study results have so far not been integrated into simulation-based research. This lack of integration stems from the absence of data that measure preferences towards PAVs and SAVs at the neighborhood level. Existing preference studies usually investigate adoption likelihood without collecting appropriate information to link preferences to precise locations or neighborhoods. This study develops a microsimulation approach, incorporating machine learning and population synthesizing, to fill this data gap, leveraging a national AV perception survey (NAVPS) and the latest National Household Travel Survey (NHTS) data. The model is applied to San Francisco, CA, and Austin, TX, to test the concept. We validate the proposed model by comparing the spatial distributions of synthesized ride-hailing users and observed ride-hailing trips. High correlations between our synthesized user density and empirical trip distributions in two study areas, to some extent, verify our proposed modeling approach.