Long automated driving phase affects take‐over performance

Abstract

We investigated the impact of automated driving phases of different durations (long vs. short) on take-over performance and driver state. Thirty participants drove on a dynamic simulator under autonomous mode for three successive periods of automated driving: a short (10 min), a long (1 hour) and another short (10 min) period, each ending with a take-over request. They performed a non-driving task, watching a film of their choice, throughout the autonomous phases. Driving performance -reaction time and quality -and driver drowsiness were assessed at each take-over. One hour of automated driving affected the driver's behaviour, leading to poorer take-over performance (longer reaction time and sharper avoidance manoeuvre) and increased drowsiness compared to a shorter autonomous period. Results also suggest that sequencing the autonomous phase in several short periods should improve the driver's take-over performance and help prevent drowsiness. Our findings sound a warning on the risks associated with long phases of automated driving. In particular, we provide evidence that series of short periods of automated driving are preferable to a long, continuous automated driving phase.

Estimating the out-of-the-loop phenomenon from visual strategies during highly automated driving

“…The results also suggest that the OOTL phenomenon took some time to appear and that it increased with the duration of automated driving (Körber et al, 2015;Feldhütter et al, 2017;Bourrelly et al, 2019).…”

“…In the aviation field, Molloy and Parasuraman (1996) showed that monitoring performance decreased with time. In the automated driving context, some studies have indicated that prolonged periods of automated driving rendered drivers further OOTL (Körber et al, 2015;Feldhütter et al, 2017;Bourrelly et al, 2019). For example, Bourelly et al (2019) observed longer reaction times (+0.5 s) to a critical event after 1 h of automated driving, compared to reactions to the same event after 10 min.…”

Dynamic scan paths investigations under manual and highly automated driving

“…Obvious limitations are that the task was only carried out in a simulator, so correspondence to real world behavior remains open, and that the task only lasted for a relatively short amount of time, so effects of learning or boredom remain uncertain. Future experiments are also required to better describe the automation supervision processes, in particular to describe how automation is impacting visual sequences over longer periods of time 65 . Investigations of visual sequences engaged in other driving scenarios would also be of interest.…”

Automation Aftereffects: The Influence of Automation Duration, Test Track and Timings

“…A period of automated driving may lead to a reduced driving performance compared to manual driving [7], [11], [12]. In addition, some evidence exists on that automation duration (i.e., the time drivers are exposed to automation) also reduces driving performance after a TOR [13], [14]. That is, longer automation duration seems to result in more degraded performance (e.g., longer response times, greater number of uncontrolled maneuvers, greater accelerations) than shorter automation duration [13], [14].…”

“…In addition, some evidence exists on that automation duration (i.e., the time drivers are exposed to automation) also reduces driving performance after a TOR [13], [14]. That is, longer automation duration seems to result in more degraded performance (e.g., longer response times, greater number of uncontrolled maneuvers, greater accelerations) than shorter automation duration [13], [14]. However, at least one study showed no effects of automation duration on driving performance [15].…”