Typology of Behavioral and Emotional Reactions to Uncomfortable Automated Driving Operations

Abstract

Driving comfort is considered one of the core factors for broad public acceptance of automated driving. Monitoring emotional and behavioral reactions to potentially uncomfortable automated driving maneuvers could allow for early interventions to avoid discomfort, e.g. by adapting the automated driving style or information presentation. In a driving simulator study, 74 participants balanced in gender and age (51% male, 19 to 75 years) were instructed to answer emails on a laptop placed at the center console during a highly automated drive. After several kilometers, they experienced a rather fast and uncomfortable approach to a stationary truck at the rear end of a traffic jam. Behavioral (take-over, glances, interruption of laptop work) as well as emotional reactions (facial expression analysis using Visage FaceTrack and FaceAnalysis v9.0) were assessed 200m before reaching the end of the traffic jam and compared to a 200m baseline. To consider individual differences, a clustering approach was applied, resulting in a typology of five reaction patterns. Cluster 1 (“not noticed”, 9%) did not interrupt the laptop work and showed no glances ahead to the approach situation. Cluster 2 (“quick check”, 15%) interrupted the laptop work only briefly but did not take the hands off the keyboard, quickly checked the situation (9.5% glance time ahead) and showed a small average peak increase in the emotion “surprise” of 4.8% compared to the baseline. Cluster 3 (“observation”, 30%) interrupted the laptop work by removing the hands from the keyboard, observed the situation (20.6% glance time ahead) and showed an increase in average peak surprise by 9.7%. Cluster 4 (“quick take-over”, 31%) observed the situation (45.1% glance time ahead), interrupted the laptop work by grasping the steering wheel, started braking rather quickly at the last moments of the approach and showed an increase in average peak surprise by 9.2%. Cluster 5 (“planned take-over”, 15%) observed the situation intensively already at a very early stage (64.3% glance time ahead), resumed manual control in a planned manner and showed little increase in average peak surprise by 3.8%. To conclude, behavioral and emotional reactions to an identical uncomfortable automated approach maneuver differ considerably between participants. Thus, information and prevention strategies to avoid discomfort cannot be designed as a one-fits-all solution, but need to be tailored to the actual state and behavior of each driver.

Keywords: Face Tracking, Emotion Recognition, Automated Driving, Driving Simulator

DOI: 10.54941/ahfe1005204