The influence of alcohol and automation on drivers’ visual behavior during test track driving

Artikel i vetenskaplig tidskrift, 2023

Background: Driving under the influence of alcohol severely increases crash risk. Impairment detection during driving is therefore key to improve traffic safety. However, future detection systems need to capture impairment for more reasons than alcohol intoxication and must also function in all driving modes (manual, assisted, autonomous). Driver Monitoring Systems (DMS) are promising candidates for such broader impairment detection.
Method: A test track study investigated the effects of alcohol intoxication on drivers’ visual behavior both when just driving and when engaged in a non-driving related task. Twenty-six participants performed two drives: 1) sober baseline, 2) with a target blood alcohol concentration of 0.1%. The participants drove in either manual, assisted, or autonomous drive (AD) mode.
Results: Intoxication influenced glance behavior in all driving modes. It was most evident during visually demanding non-driving related tasks where it resulted in longer single and total off-path glance durations. Additionally, when just driving in manual mode, almost one third of the drivers displayed gaze concentration to the forward roadway when intoxicated. For sober driving, the difference in visual behavior between manual and assisted mode were moderate. In contrast, there was a huge shift towards longer off-path glances, lower percent road center, and lower off-path glance frequency in AD mode. Conclusions: Intoxication clearly affects drivers’ on/off road glance behavior. However, it is necessary to account for both driving mode and engagement in non-driving related tasks to reliably distinguish sober from drunk driving. Glance metrics has the potential to serve as a sub-set of indicators for a broader DMS-based detection of impaired driving, which can inform the decisions on when to activate in-vehicle countermeasures.