Automated driving functions, the decisions they make and how they are understood - Acceptance, trust and perceived safety from the user's perspective

Aaron Lehel Baltazar Edelmann

AutorIn

Aaron Lehel Baltazar Edelmann

Titel

Automated driving functions, the decisions they make and how they are understood - Acceptance, trust and perceived safety from the user's perspective

Zitierfähige Url:

https://nbn-resolving.org/urn:nbn:de:bsz:14-qucosa2-1018245

Erstveröffentlichung

2026

Datum der Einreichung

22.05.2024

Datum der Verteidigung

07.02.2025

Abstract (EN)

In reference to IEEE copyrighted material which is used with permission in this thesis, the IEEE does not endorse any of Technische Universität Dresden's products or services. Internal or personal use of this material is permitted. If interested in reprinting/republishing IEEE copyrighted material for advertising or promotional purposes or for creating new collective works for resale or redistribution, please go to http://www.ieee.org/publications_standards/publications/rights/rights_link.html to learn how to obtain a License from RightsLink.

Throughout the past years, automated driving has been one of the central themes in the automotive industry. As a result, numerous advanced driver assistance systems, such as lane and distance keeping assistants, are increasingly available and being implemented in modern vehicles. What is more, first conditional automated functions, which allow temporary full control by the automated system of all aspects of the driving task without constant monitoring, are now being introduced. This advancement is anticipated to revolutionise road traffic, offering benefits such as fewer crashes and reduced fuel consumption, increased traffic efficiency, and driver relaxation and engagement with non-driving activities. However, there are major concerns regarding the public's willingness to purchase and use automated vehicles. In this regard, it is assumed that psychological barriers will pose challenges to realising the full potential of automated systems. Without overcoming these barriers, automated vehicles will not be able to tap their potential. Hence, researching and understanding factors impeding their introduction and usage is therefore a key to designing a successful system. The present thesis aims at evaluating automated driving functions with regard to two main barriers preventing successful introduction. Firstly, it is important to ensure users have an appropriate understanding of automated driving systems and to find means of supporting them in gaining a good understanding. Within this field of action, this work especially focused on the evolution of acceptance, trust and perceived safety with increasing usage. Then, secondly, it is important to design characteristics of automated driving systems in such a way that they foster acceptance as well as trust and that they are perceived as safe. Here, this work particularly concentrated on decisions at the manoeuvre level. Moreover, as part of the second field of action, the evolution of perceived safety in relation to a function's characteristics was analysed. Within this work, six studies were conducted in light of the above-mentioned two fields of actions: Study I focused on current advanced driver assistance systems and their users' understanding of them as well as the effect this had on acceptance, trust and perceived safety. In an online survey, users answered questions on one of three advanced driver assistance systems (automated speed and distance keeping assistant, lane-keeping assistant or parking assistant) regarding their perceived understanding as well as their actual, objective understanding. Furthermore, they rated their acceptance, trust and perceived safety. The results revealed a discrepancy between perceived understanding, which was rather high, and objective understanding, which was rather low. Furthermore, acceptance, trust and perceived safety was rather high, implying that users of such functions have inadequate levels of acceptance, trust and perceived safety. Thus, it is necessary to support current and future users in order to achieve an appropriate understanding regarding automated systems abilities and limits. Study II aimed at analysing a way of supporting potential automated driving function users prior to initial usage. It investigated the impact of text-based versus video-based instructions. To assess the effects of these instruction methods, the development of acceptance, trust and perceived safety was examined during various stages of system interaction with an automated parking assistant. Therefore, an on-road study during which participants completed ten parking manoeuvres was carried out. Initially, participants in the video-based instruction group exhibited higher ratings of acceptance and trust compared to those in the text-based instruction group. Moreover, acceptance, trust and perceived safety increased with continued system usage. With increasing usage, the evaluations from both groups converged and showed no significant differences after ten parking manoeuvres. Nevertheless, the video-based instruction method demonstrated potential in mitigating initial distrust and reluctance towards automated driving functions and therefore should be considered as an instructional approach. Study III examined the impact of specific versus non-specific feedback on the evolution of acceptance and trust with a level 2 automated parking function. It, too, aimed at understanding how acceptance and trust evolve and how the development differs between feedback specificity levels. Here, an on-road longitudinal study was conducted. For the first twenty manoeuvres, no specific feedback was provided for faults due to system limitations. In the subsequent twenty manoeuvres, specific feedback addressing situational factors and system constraints was given. The findings indicated that non-specific feedback failed to foster acceptance and trust as errors stemming from system and sensor limitations remained unexplained and unpredictable. Conversely, specific feedback tailored to the situation facilitated the formation and stabilisation of acceptance and trust. This aligned with the power law of learning. Consequently, it is recommended that users receive specific feedback to understand system limitations, thereby promoting trust and acceptance development and ensuring appropriate system utilisation. Study IV addressed the evolution of perceived safety as well as the multifaceted impact that a single system characteristic can have on the perception thereof. Seven safety distances of an automated parking system were implemented resulting in parking manoeuvres of varying efficiency (in terms of required moves). Participants experienced each configuration twice and then provided ratings for perceived safety and perceived usefulness. A gradual increase in perceived safety could be observed for some participants, while others exhibited a threshold effect. Additionally, the findings indicated that prioritizing maximum safety distances led to high perceived safety, but also decreased perceived usefulness due to reduced efficiency. Conversely, prioritizing maximum efficiency resulted in lower perceived safety and, consequently, reduced ratings of perceived usefulness. Study V concentrated on the effects of a highway assistant function's driving decisions depending on the cultural background. To evaluate the impact on acceptance, an online experimental study was carried out across China, Germany, Japan, and the US. Participants were presented four overtaking scenarios, each described briefly, in which an automated vehicle made a decision. These scenarios varied in terms of the action taken (overtaking or staying in lane) and the potential consequence for another driver (low or high hindrance). Subsequently, participants provided ratings of their acceptance levels. The findings revealed that acceptance is contingent upon the driving decisions made and is additionally influenced by participants' cultural backgrounds. Study VI took up on the previous two studies and analysed overtaking manoeuvres of an automated vehicle in a driving simulator with regard to acceptance, trust and perceived safety. Five different safety distances were implemented and examined in six conditions, which differed in terms of velocity, costs involved in foregoing the overtaking manoeuvre and the perspective from which the overtaking manoeuvre was experienced (from within the automated vehicle or in a vehicle approaching from behind). The results suggest that with increasing safety distance, the automated vehicles' overtaking decisions are deemed more acceptable, more trustworthy and safer. Furthermore, with higher velocity and higher costs involved smaller safety distances became more acceptable, more trustworthy and were perceived to be safer. In addition, the same manoeuvre decisions were assessed less acceptable and the AV was perceived as less trustworthy and less safe when viewed from the perspective of the vehicle approaching from behind. Besides investigating the topics in the studies individually, the findings and their theoretical as well as practical implications are discussed more broadly. This includes the merits and demerits of overly enthusiastic and self-assured users, the ambiguity of perceived safety as well as the resulting inferences for driver modelling, and finally, the challenges involved in optimising automated driving functions. In summary, the potential of automated driving systems is enormous. To this day, it has not yet been fully attained. By exploring different facets that hinder adequate usage and successful adoption, this work contributes to overcoming some of the wide range of barriers that have to be addressed in order for automated driving systems to be effective and advantageous to users, the transportation sector and society in general.

Verweis

Veröffentlichung, auf die sich das Kapitel 6 der Dissertation bezieht
Specific Feedback Matters - The Role of Specific Feedback in the Development of Trust in Automated Driving Systems
DOI: 10.1109/IVS.2019.8814126

Effects of User Instruction on Acceptance and Trust in Automated Driving
Veröffentlichung, auf die sich das Kapitel 5 der Dissertation bezieht
DOI: 10.1109/ITSC45102.2020.9294511

Cross-cultural differences in the acceptance of decisions of automated vehicles
Veröffentlichung, auf die sich Kapitel 8 der Dissertation bezieht
DOI: 10.1016/j.apergo.2020.103346

The interaction between perceived safety and perceived usefulness in automated parking as a result of safety distance
Veröffentlichung, auf die sich das Kapitel 7 der Dissertation bezieht.
DOI: 10.1016/j.apergo.2022.103962

Freie Schlagwörter (DE)