Nicole
Dillen,
Master’s
candidate
David
R.
Cheriton
School
of
Computer
Science
Despite rapid advancements in automated driving systems (ADS), current HMI research tends to focus more on the safety driver in lower level vehicles. That said, the future of automated driving lies in higher level systems that do not always require a safety driver to be present. However, passengers might not fully trust the capability of the ADS in the absence of a safety driver. Furthermore, while an ADS might have a specific set of parameters for its driving profile, passengers might have different driving preferences, some more defensive than others. Taking these preferences into consideration is, therefore, an important issue which can only be accomplished by understanding what makes a passenger uncomfortable or anxious.
In order to tackle this issue, we ran a human study in a real-world autonomous vehicle. Various driving profile parameters were manipulated and tested in a scenario consisting of four different events. Physiological measurements were also collected along with self-report scores, and the combined data was analyzed using Linear Mixed-Effects Models. The magnitude of a response was found to be situation dependent: the presence and proximity of a lead vehicle significantly moderated the effect of other parameters. Stopping events also generated a higher level of response than non-stopping events. Finally, a statistically significant association between physiological responses and self-reported scores showed that such responses could potentially be used to indicate comfort or anxiety in future adaptive systems.