As research for the development of autonomous driving technology is
actively progressing, the commercialization of level 3 conditional
autonomous vehicles is in full swing.
Conditional autonomous vehicles refers to vehicles in which the vehicle
syste...
As research for the development of autonomous driving technology is
actively progressing, the commercialization of level 3 conditional
autonomous vehicles is in full swing.
Conditional autonomous vehicles refers to vehicles in which the vehicle
system recognizes all overall driving situations and performs stable
autonomous driving under specific road conditions. At this time, the driver
deviates from vehicle control and driving responsibilities, and will be
guaranteed freedoms for doing Non-Driving Related Task(NDRT), such a
s using smartphone, watching media videos and etc. However, since
conditional autonomous vehicles can generate Take-Over Request at any
time, the driver must always exist as a ‘fallback-ready driver’ that can
safely continue manual driving. And for this, the existence of an AI agent
that supports driver is essential even in situations where safe autonomous
driving is being carried out.
This study explored the effective information provision method of
interactive AI agents that support drivers, focusing on the driver’s
situational awareness ability, which is deteriorated due to the performance
of NDRT in conditional autonomous driving environments. Specifically, it
was investigated through experiments how the AI agent’s level of
explainability and visual cues affect driver’s trust, perceived safety,
context awareness, perceived interruption, intention to use, and situation
awareness.
As a result, it was found that both the high level of explainability and
the case where visual cues are provided showed positive effects on the
driver’s trust, perceived safety, context awareness, intention to use, and
situation awareness. However, in terms of perceived interruption, both
high levels of explainability and visual cues were found to have negative
effects that hinder the driver’s consistent flow of NDRT.
This study is significant in that it presented a design guideline for an
effective interaction method between a driver and a vehicle based on a
conditional autonomous driving environments