Citation

Ziakopoulos A, Telidou C, Anagnostopoulos A, Kehagia F, Yannis G. IATSS Res. 2023; 47(4): 499-513.

Copyright

(Copyright © 2023, International Association of Traffic and Safety Sciences, Publisher Elsevier Publishing)

DOI

10.1016/j.iatssr.2023.11.002

PMID

unavailable

Abstract

The present research investigates a range of factors affecting autonomous vehicle (AV) acceptance of Greek citizens through a questionnaire distributed to 563 respondents. Following the extraction of descriptive statistics, self-organizing maps (SOMs) were employed to meaningfully categorize and aggregate questions pertaining to four main pillars of the questionnaire, which are conceptually relevant namely: (i) how several factors affect general car choices of respondents, (ii) what the respondents perceived that AVs would offer, (iii) how much they agreed with stated expected technology and efficiency-oriented AV traits and (iv) how they believe several factors affect driving behavior overall. A Random Forest (RF) algorithm was applied to classify the AV acceptance decisions of a training subset of the respondents, and was subsequently assessed on a test subset. SOM results indicate that participants can be meaningfully separated into two SOM cluster groups for pillars (i), (ii) and (iv), while pillar (iii) yielded separations into three SOM cluster groups. RF feature importance calculation indicated a number of affecting variables; the five most contributing ones are: distance covering capabilities of AVs was a major factor affecting acceptance decisions, followed (by a wide margin) by responder opinions on whether the principles and conscience of drivers can be replaced by an AI navigator without reducing safety levels, while the algorithm itself conducted successful classification to about 80% of test cases. Present results can be used to anticipate AV penetration levels based on sample characteristics and to improve AV traits in cases where higher AV penetration is sought.

Language: en

Keywords

Automated vehicles; Information mining; Perception analysis; Random Forest; Self-Organizing Maps