Supporting Drivers in Take-over Situations in Highly Automated Driving

06. Juli 2017, 13:00 , 14:30

Veranstalter:  Shadan Sadeghian Borojeni Universität Oldenburg
Ort:  OFFIS, Escherweg 2, Raum F02

Abstract:
With the rise of advanced driving assistant systems (ADAS), highly automated driving has become more
probable. According to NHTSA, the next generation of automated vehicles on the road belong to level 3 of
automation which is "Limited Self Driving Automation". In this level, the vehicle is fully automated but the
driver has to be available to take over control in cases of hazards with sufficient transition times. These
situations are called take-over situations. In take-over situations, drivers have to switch from their secondary
task to the driving task. During this switch, they have to shift their attention from one task to the other, which
requires perceiving the state of the driving environment, make decisions, and act accordingly. Being engaged
with secondary tasks can cause drivers to be vulnerable to delays or errors when getting back to the driving
task due to not having a chance for attending to it and not being in the loop. This can lead to hazardous
situations if the driver has to take over vehicles control due to automation shortcomings. Therefore, appro-
priate user interface designs for Take over Requests (TOR) are required to ensure smooth transitions from
secondary to driving task.
Despite that advances in computing have allowed users to perform multiple concurrent activities or switch
between tasks, when interacting with machines, human cognitive capabilities have not increased, leaving us
vulnerable to errors. Therefore, priming drivers for these transitions can be highly beneficial. This thesis
investigates how contextual information can be conveyed through multi-modal cues to the drivers to prime
them with the information about take-over situation and assist decision making. This information consists of
steering direction, location of traffic objects, road curvature, and recommended manoeuvres. We pursue the
effect of the following factors in design of TORs on drivers' performance in take-over situations: a) presen-
tation modality, design, and patterns, b) motion perception and understanding situation urgency, c) drivers'
engagement in non-driving tasks, and d) reliability of TORS. We designed and implemented multi-modal
TORs and integrated them in driving simulators. These TORs were evaluated in several studies conducted in
fixed and motion-based driving simulators in which drivers’ responses to contextual cues are measured. We
applied different approaches such as using desktop vs. high fidelity simulators, and using different cognitive
tests for the non-driving activities.
In conclusion, this thesis shows that applying multi-modal cues that convey information about driving context
as TORs, can improve drivers' performance in take-over situation despite their high engagement in non-driv-
ing tasks, or situation urgency. It contributes design guidelines for development of the future vehicles, where
the automation level varies from partly, to highly or fully automated, and the driver-out-of-the-loop problem
rises. The results of this thesis, can be a motivation for future research and a support for industries in devel-
oping assistant systems for drivers to increase their situation awareness in cases of hazards.
Betreuerin: Prof. Dr. Susanne Boll