Building Pedestrians' Trust and Awareness while Interacting with Automated Vehicles: A Solution through e-HMI Designs

Abstract

The introduction of high-level automated vehicles (AVs) into the current transportation system is expected to enhance road safety, including that of vulnerable road users (VRUs). However, one challenge to their deployment is the lack of typical communication cues exchanged between human drivers and pedestrians. One potential solution to replace those human-to-human cues is the development of external human-machine interfaces (e-HMIs).Although several researchers have introduced e-HMI concepts, there remains a deficit in guidance for optimizing these designs. In designing effective and efficient e-HMIs, it is important to incorporate peopleâs needs by considering all stakeholders and reaching out to a representative sample. Moreover, while establishing appropriate levels of trust among pedestrians upon their interaction with AVs is crucial, there has been scant research into the potentially adverse effects of varying trust levels. A multitude of factorsâincluding pedestriansâ demographics, behaviors, and familiarity with AVs, AVs' driving styles, and crossing locationsâcontribute to the spectrum of trust in AVs. This spectrum ranges from overtrust to distrust, each posing distinct implications for pedestrian safety and mobility. To mitigate risks associated with such trust levels, it is vital to engineer e-HMIs that facilitate safe and efficient interactions. To bridge the identified research gaps, this dissertation provides a novel, structured approach to crafting safe and effective e-HMIs, utilizing a three-phase framework: (1)Formulation, (2)Derivation, and (3)Validation. In the Formulation phase, the study conducts an in-depth analysis of historical vehicle-pedestrian crash data. The analysis revealed an increased risk associated with straight-moving vehicles to pedestrians at intersections, highlighting the growing vulnerability of older pedestrians in recent years. By quantifying these factors, the study identifies key pre-crash variables essential to developing AV-pedestrian use cases.During the Derivation phase, the work seeks to understand user needs through participatory design, engaging both experts in the domain and a cross-section of the general public via surveys. Findings indicated that pedestriansâ e-HMI preferred features were heavily shaped by their current interactions with human-driven vehicles (HDVs), and these preferences were influenced by their personal attitudes and demographics. Expert feedback also unearthed distinct preferences and highlighted specific challenges tied to certain e-HMIs.In the Validation phase, the research embraces textual and symbolic e-HMI prototypes informed by earlier survey insights. These prototypes' influence on pedestriansâ trust, situation awareness, and cognitive workload was assessed in a virtual reality setup. Through collecting and analyzing self-reported, behavioral, and physiological measures, the results underscore the potential for symbolic e-HMIs to mitigate pedestrian distrust and frustration, while also bringing attention to potential unintended consequences.This dissertation presents an extensive approach to improving pedestrian safety and mobility, through applying an integrated data-driven and user-centered e-HMI design process. Consequently, it advances our comprehension of pedestrian requirements when engaging with AVs, enabling the derivation of more precise potential interaction scenarios, and paving the way for more customized e-HMI configurations. This work informs the development of novel e-HMI guidelines aimed at fostering optimalânot merely maximalâlevels of pedestrian trust in AVs at both controlled and uncontrolled crossing locations. By fine-tuning the level of trust, the study enhances the public's receptivity towards AVs. Moreover, the outlined framework holds potential for wider application, including developing and evaluating HMI solutions across various industries. The anticipated outcome is a more seamless integration of automated systems into daily life, predicated on human-centered design principles prioritizing both safety and trust.