The Sooner the Better: Drivers’ Reactions to Two-Step Take-Over Requests in Highly Automated Driving

Abstract

Driving behavior after take-over requests (TORs) is one of the most popular subjects in human factors re-search on highly automated driving. Many studies utilized one-step TOR procedures to prompt drivers to resume vehicle control. The present paper examines driver behavior when experiencing a two-step TOR procedure in different modalities. A two-step TOR gives drivers a choice to resume vehicle controls be-tween a warning (first step) and an alarm (second step). Our findings indicate that a substantial number of drivers resumes vehicle controls after the second step, resulting in a higher number of crashes. More generally, criticality of the driving situation increases with increasing reaction times. Driving and interview data suggest that step two of the TOR should be presented earlier. Alternatively, a multi-step TOR could be used to increase drivers’ situational awareness. Auditory TORs are associated with shorter reaction times than visual-auditory TORs. Implications on TOR design are discussed.

Behavioral Changes to Repeated Takeovers in Highly Automated Driving: Effects of the Takeover-Request Design and the Nondriving-Related Task Modality

“…This resulted in 171 data points in the first analysis interval and 216 trials in the second analysis interval. For a more detailed analysis of the takeovers, see Epple, Roche, and Brandenburg (2018). Even though all participants took over control, in fourteen trials (6.5%), participants collided with the construction site or drifted from the road.…”

The TOR Agent:Optimizing Driver Take-Over with Reinforcement Learning

“…Compared to imminent situations, scheduled transitions allow for planning within a certain time window. For example, recent works have looked into pre-alerts ("priming" [24]) and multi-stage warnings [5] to improve drivers' TOR responses. However, not many works have addressed the question of when exactly a user interface should communicate a TOR to the driver.…”

“…Our approach is novel in a way that it aims to determine the potentially best timing to communicate TORs to drivers. Such a system could become a valuable extension for in-vehicle interfaces, e.g., for multi-stage TORs [5]. Although the challenge of determining an appropriate timing to interrupt drivers has been proposed by empirical works both in the context of side activities [22] and transitions [28], we are not aware of any project investigating how such interfaces could be implemented.…”

Assessing the Effect of Countdown Featured TOR Signal on Drivers in Automated Driving Mode Change

“…Therefore, there have been substantial numbers of studies to design effective TOR signals. Regarding the channel or modality of the signal, for example, Epple et al and Walch et al revealed that drivers preferred a combination of visual, auditory and tactile cues for the TOR signal rather than the use of a single channel (Epple et al, 2018;Walch et al, 2015). Moreover, a combination of generic sounds and speech sounds can trigger the fastest drivers' response time (Forster et al, 2017).…”