Loughborough University
Browse

Understanding automated vehicles communication of yielding intentions to pedestrians in complex urban traffic environments

Download (39.44 MB)
thesis
posted on 2024-10-25, 14:25 authored by Elena Fratini

Automated vehicles will soon be introduced in our existing transport system, giving rise to a mixed traffic environment. One of the anticipated outcomes of this transition will be the requirement for automated vehicles to be able to interact with pedestrians, replacing the current informal driver-pedestrian communication strategies. To this end, external human-machine interfaces (eHMIs) are being investigated in different shapes and forms to ease cooperation with automated vehicles and help build pedestrians’ trust towards these new technologies. eHMIs can be used to communicate various messages which may benefit a smooth interaction, importantly concerning the vehicle’s intentions e.g. whether the AV intends to stop to give way or “yield” to a pedestrian. The communication of the vehicle’s intentions has the benefit of making clear to the pedestrian what the vehicle is about to do, thus ensuring a more transparent interaction and setting the right expectations.

A key aspect of effective eHMIs is that they should promote safety and in order to do so they should be designed to perform as expected and enable the timely communication of supporting information while leaving the crossing decision in the hands of the pedestrian i.e. avoid impairing their exploratory behaviour when crossing. Research has shown that eHMIs can be misleading and can lead to early crossing decisions. However, the full operational extent of eHMIs is currently limited to a restricted range of simple testing scenarios, which do not reflect the variety and complexity of the real world where eHMIs will be ultimately deployed.

The overall goal of this thesis is to understand how eHMIs are perceived and reacted to in different traffic scenarios with different characteristics and levels of complexity resembling more realistic urban traffic environments. A web-based survey (n = 96) was initially conducted to investigate whether road priority, eHMI animation pattern and eHMI activation distance (i.e. vehicle proximity to the crossing point when the eHMI is activated), impacted eHMI message intuitiveness as an intention or an instruction message. The results showed that eHMI messages can be interpreted as instruction orientation messages for different road, eHMI animation and distance attributes. The results also showed that a vehicle’s proximity (or eHMI activation time) and kinematics can be important cues for pedestrians when deciphering an AV’s intentions, in other words, that the context in which a signal is shown can be decisive in interpreting its meaning. To additionally expand on and to determine the contextual elements which might impact pedestrians’ crossing behaviour and successful interactions with automated vehicles and eHMIs, a small-scale (n = 15) study was conducted to identify a range of indicators that contribute to the crossing scenario complexity and then rank these by increasing complexity. A set of crossing scenarios was thus developed and tested for this purpose in a simulated virtual reality environment. Objective crossing behaviour measures were collected. Sense of presence and simulator sickness were also recorded as a measure of the ecological validity of the virtual environment.

The results indicated that some crossing scenarios were more taxing for pedestrians than others, such as those with road geometries where traffic approached from multiple directions. Further, the presence scores showed that the virtual environments experienced were found to be realistic. These findings provided a base for further developing the traffic simulated environments to test eHMIs under increasingly challenging crossing circumstances, which were used in a subsequent virtual reality study. Trials (n = 19) were conducted where different eHMIs and eHMI activation times were tested in three different levels of scenario complexity i.e. a “low complexity” one lane road with one-way traffic, a “medium complexity” environment consisting in a junction layout with two-lane, two-way traffic, and a “high complexity” scenario with the same road geometry as the previous, but with poor weather conditions, the presence of visual and cognitive distractions and a time pressure task. The findings showed that the presence of the eHMI had an impact on pedestrian behaviour and attention. Complexity of the crossing scenario impacted pedestrians’ perceived time pressure, elements of the situation awareness, and brought to light risky pedestrian behaviours.

The findings of this research provide new research to inform pedestrian-AV interactions. The main implications are as follows:

• The VR environment is an effective tool for the development and testing of autonomous systems with a human in the loop.

• Complex scenarios are essential to develop resilient eHMI systems and safe pedestrian-AV interactions.

• Certain eHMI characteristics may inadvertently reduce pedestrian’s attention to traffic, potentially increasing safety risks.

• Light-based eHMI messages can be ambiguous and interpreted equally well as a vehicle’s intention to yield to pedestrians and instructions for pedestrians to cross.

• Vehicle kinematics play a significant role in pedestrian-AV interactions.

• Pedestrians tend to exhibit bimodal crossing behaviours when interacting with AVs even in more complex traffic scenarios, typically crossing either when the vehicle is over 38.5 meters away or when it has stopped. In complex scenarios the gap offered by the AV for the pedestrian to cross can be rejected.

• Over-trust towards the AV has the potential to cause risky crossing behaviours while under-trust can result in missed crossing opportunities.

History

School

  • Design and Creative Arts

Department

  • Design

Publisher

Loughborough University

Rights holder

© Elena Fratini

Publication date

2024

Language

  • en

Supervisor(s)

Ruth Welsh ; Pete Thomas

Qualification name

  • PhD

Qualification level

  • Doctoral

This submission includes a signed certificate in addition to the thesis file(s)

  • I have submitted a signed certificate

Ethics review number

2020-2378-2052; 2021-6343-6447; 2023-13088-12951