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

Abstract

Automation aftereffects (i.e., degraded manual driving performance, delayed responses, and more aggressive avoidance maneuvers) have been found in driving simulator studies. In addition, longer automation duration seems to result in more severe aftereffects, compared to shorter duration. The extent to which these findings generalize to real-world driving is currently unknown. The present study investigated how automation duration affects drivers' take-over response quality and driving performance in a road-work zone. Seventeen participants followed a lead vehicle on test track. They encountered the road-work zone four times: two times while driving manually, and after a short and a long automation duration. The take-over request was issued before the lead vehicle changed lane to reveal the road-work zone. After both short and long automation durations, all drivers deactivated automation well ahead of the road-work zone. Compared to manual, drivers started their steering maneuvers earlier or at similar times after automation (independently of duration), and none of the drivers crashed. However, slight increases in vehicle speed and accelerations were observed after exposure to automation. In sum, the present study did not observe as large automation aftereffects on the test track as previously found in driving simulator studies. The extent to which these results are a consequence of a more realistic test environment, or due to the duration between the timings for the take-over request and the conflict appearance, is still unknown.

Driver Response to Take-Over Requests in Real Traffic

“…These probability distributions can be used by system designers to select values to match their priorities and understand the uncertainty in their decisions. For example, if a TOR is designed to be issued 7 s prior to exiting the ODD, we cannot be sure that all drivers will manage to deactivate automation in that time, as the 95% HPD for the 95th percentile driver who is looking towards an NDRT item when the TOR is issued ranges between 5.1 s and 8.0 s. Despite the advantages of the Bayesian framework, it is rarely used in the literature on human factors of automated driving, with a few exceptions (e.g., [26], [27], [28]).…”

Driver Visual Attention Before and After Take-Over Requests During Automated Driving on Public Roads

“…The expectation that drivers can respond appropriately to TORs by safely resuming manual control has been questioned by human factors researcher ( Gold et al, 2013 ; Louw et al, 2015 ; Seppelt & Victor, 2016 ). To date, drivers’ responses to TORs have been explored mainly in driving simulators ( Gold et al, 2013 ; Louw et al, 2015 ; McDonald et al, 2019 ), and to a lesser extent in more realistic settings such as real roads ( Eriksson et al, 2017 ; Naujoks et al, 2019 ) or test tracks ( Pipkorn et al, 2021 , 2022 ). The focus of the current research is the time needed for drivers to deactivate AD by steering, braking, or pressing buttons (the take-over time; McDonald et al, 2019 ).…”