Wednesday 9:00-10:30
„What Else Influences Customer’s Willingness on Buying an Automated Driving Vehicle? A Web-Survey Study.“
Huiping Zhou, Joseph Leung, Makoto Itoh and Satoshi Kitazaki
Abstract— This paper is focused on user’s attitudes to driving automation (AD) on the basis of their mastered/seeking knowledge in order to reveal influences of the AD’s knowledge on customer’s willingness on buying automated driving vehicles (AVs). An online survey was conducted for 1072 participants who were classified into 7 groups in terms of their attitudes to AD and intentions of buying AVs. According to the results, it is found that users’ mastering knowledge affected their decision-making to purchase AVs, and potential buying willingness also impacted what extent of the seeking knowledge. Additionally, a gap between purchasers’ expectation and sellers’ instruction was indicated, and became greater when customer had a high unwillingness.
„The Influence of Multiple Cars on Pedestrian Behavior During Continuous Road Crossing Investigated in an Omnidirectional Walking Simulator in VR“
Max Theisen, Melina Bergen, Wolfgang Einhäuser and Caroline Schießl
Abstract— In recent years, many authors have made the case for external human-machine interfaces (eHMIs) for automated vehicles (AVs). These should replace the communication channels of the driver who has become redundant in AVs (see [Dey et al., 2020] for an overview). It is crucial to first understand pedestrians’ perception, decision-making and action in today’s non-automated traffic in order to appropriately design eHMIs for future pedestrian-AV interaction. One important scenario is a pedestrian wanting to cross a road in front of an oncoming vehicle. In this scenario, previous studies often make two simplifications: (1) Firstly, it is typically assumed that pedestrians are stationary at the beginning of the interaction with a vehicle ([Kalantarov et al., 2018], [Sobhani and Farooq, 2018 ([Kooijman et al., 2019]). However, it is known from traffic observations that pedestrians rarely stop unnecessarily at the curb, but rather start the search process for vehicles already while walking towards the curb ([Gorrini et al., 2018]). In most cases, pedestrians decide whether or not to cross the road by either accelerating or decelerating before reaching the curb. We address this issue by studying how pedestrians approach a crossing location while looking for crossing opportunities at the same time. Technically, this is made possible by using treadmill systems such as the Omnideck ([Brinks and Bruins, 2016], an omnidirectional treadmill that enables continuous walking in VR. (2) Secondly, so far few studies have investigated two-lane crossings with traffic from both sides ([Neider et al., 2010] and [Mok et al., 2022] being exceptions), although two-lane roads are the most common type of road in most places. We address how visual attention is distributed to both vehicles and how the gap-acceptance decision is influenced by a second car simultaneously approaching from the other direction with a certain time gap to the first vehicle. In our study, we investigate the road crossing behavior of 24 participants during continuous walking across two-lane roads. We compare pedestrians’ visual attention (head- and eye-movements) and gait (walking trajectory and velocity) for single- and dual-vehicle trials for different velocities and time-gaps of the two vehicles. The results are likely to have implications for the design of eHMIs for AVs as well as for predicting pedestrian behavior during interactions with AVs.
Keywords: Pedestrian-vehicle interaction, Vulnerable road users, Pedestrian crossing, Virtual reality, Omnidirectional treadmill, Gait, Head-movements, Eye-tracking
„Differences in Pedestrian Behavior at Crosswalk Between Communicating With Conventional Vehicle and Automated Vehicle in Real Traffic Environment“
Masahiro Taima and Tatsuru Daimon
Abstract— In this study, we examine the differences in pedestrian behavior at crosswalks between communicating with conventional vehicles (CVs) and automated vehicles (AVs). To analyze pedestrian behavior statistically, we record the pedestrian’s position (x- and y-coordinates) every 0.5 s and perform a hot spot analysis. A Toyota Prius (ZVW30) is used as the CV and AV, and the vehicle behavior is controlled using the Wizard of Oz method. An experiment is conducted on a public road in Odaiba, Tokyo, Japan, where 38 participants are recruited for each experiment involving a CV and an AV. The participants cross the road after communicating with the CV or AV. The results show that the pedestrians can cross earlier when communicating with the CV as compared with the AV. The hot spot analysis shows that pedestrians who communicates with the CV decides to cross the road before the CV stops; however, pedestrians who communicate with the AVs decide to cross the road after the AV stops. It is discovered that perceived safety does not significantly affect pedestrian behavior; therefore, earlier perceived safety by drivers’ communication and external human-machine interface is more important than higher perceived safety for achieving efficient communication.
„May I Go First or You? The Effect of Right-Of-Way on Pedestrians’ Trust For the Interaction With Automated Vehicles“
Merle Lau and Michael Oehl
Abstract— In the near future, automated vehicles (AVs) will have to interact with other traffic participants. This leads to the challenge that AVs need to be able to communicate successfully to ensure traffic safety, in particular, with vulnerable road users, e.g., pedestrians, as they highly depend on the external communication with their surrounding traffic environment. The use of communication tools for AVs, e.g., external HMI (eHMI) and dynamic HMI (dHMI), can contribute to a well-working interaction between AVs and pedestrians. External HMIs transmit explicit communication signals, whereas, dHMIs transmit implicit communication signals via the vehicle’s dynamics. Research showed promising results for both communication tools separately and in combination when it is well coordinated. However, some studies also manifested negative effects of eHMIs, i.e., overtrust. As a limitation, this was investigated mostly in traffic scenarios where the pedestrians had the right-of-way, which might have affected pedestrians’ subjective assessment and, therefore, needs further investigation. In this experimental online study, pedestrians interacted in pre-recorded video sequences with AVs approaching from the right-hand side, i.e., from the non-privileged side. The AVs were equipped with a 360° LED light-band eHMI. Moreover, the interaction took place on a shared space traffic scenario which is defined by interactions in low-speed and low-distance and, thus, this traffic scenario not only demands implicit communication signals by the dHMI but also explicit information signals by the eHMI. In the experimental conditions, the vehicle size (car vs. bus), the vehicle’s dynamics, i.e., dHMI (yielding vs. non-yielding), and the eHMI condition (no eHMI vs. static eHMI vs. dynamic eHMI) was varied. A major focus was placed on the experimental conditions in which the eHMI contradicted the dHMI, i.e., when the dynamic eHMI showed a yielding intent by a slow pulsation, although the vehicle was not yielding (non-matching conditions). The results revealed that pedestrians‘ trust was surprisingly high in those non-matching conditions which could present a high-risk traffic scenario for pedestrians as vulnerable road users. Overall, findings of this study provided insights into possible negative effects of eHMIs when they do not match the vehicle’s dynamics.
„Vehicle Behaviour + External Human-Machine Interface = ?: Investigating Impacts of Implicit and Explicit Cues from Automated Vehicles in Japanese Parking Areas“
Jieun Lee, Jun Soshiroda, Kosuke Takizawa, Tatsuru Daimon and Satoshi Kitazaki
Abstract— With the introduction of driverless automated vehicles to parking areas, it is important to draw a clear picture of communication strategies for pedestrians-automated vehicles (AVs) interaction. This study examined how different types of cues from AVs in parking areas affected pedestrians’ behavior and attitudes toward AVs. VR simulations manipulated 4 between-subject conditions based on text messages of external human-machine interfaces (no eHMI, „After you“, „I’ll stop“, „Automated driving“) and included 6 trials considering 2 types of AVs (golf-cart, mini-bus) and 3 types of vehicle behaviors (Normal Deceleration, Early Deceleration, and Early Stop). Results showed pedestrians could not perceive that AVs tended to give way to them when AVs stop too early. Giving text cues from eHMIs with the early deceleration improved pedestrians’ trust and relieved anxiety toward AVs. These findings based on the combination of implicit and explicit communication strategies are likely to be of interest to not only system designers, but also government.
„Should eHMIs communicate an automated vehicle’s behavior or intention?“
D. Eisele, M. Schlemer and T. Petzoldt
Abstract— It has been shown that external human machine interfaces (eHMIs) have the potential to influence pedestrians’ behavior in interactions with automated vehicles (AVs) in a desirable manner. Currently, policymakers suggest that AVs should use a light-emitting eHMI to communicate their driving status (automated/human-driven). However, it has been argued that displaying driving status alone might lead to overtrust, resulting in unsafe crossing behavior in front of AVs. Recent evidence suggests that an additional eHMI could potentially mitigate overtrust and its undesirable consequences by providing higher system transparency. The objective of the resented study was to compare the effects of two different eHMI messages referring to either the AV’s behavior (“I am braking”) or its intention (“I intend to yield to you”) on pedestrians’ crossing decisions and trust in the automated system. A between-subjects laboratory study (N = 87) was conducted. Participants were shown 31 videos of approaching conventional vehicles (CVs) and AVs from the perspective of a pedestrian intending to cross the street. AVs yielded in accordance with the regulation apart from one trial in which the AV failed to yield. First results show that both the behavior and intention eHMIs led to earlier crossing decisions in front of AVs than CVs. There was no difference in trust between the two messages as long as the AV yielded in the expected manner. When it failed to yield, we observed a significant amount of potentially unsafe crossing decisions and a decrease in trust in the automated system. The intention eHMI seems to have mitigated the potentially unsafe crossing decisions better than the behavior eHMI. More detailed results will be resented. Implications for the design of eHMIs will be discussed.