Detecting unacceptable behavior of an autonomous vehicle using electroencephalography

Abstract

Abstract In the present study, we investigated event related potentials (ERPs) in the context of autonomous driving - specifically in left-turn situations through oncoming traffic. We recorded electroencephalography while participants (n = 33) observed a simulated autonomous vehicle executing a left turn maneuver through oncoming traffic. In the ERP, we observed an increased N2 (251 to 431 ms) when the AV behaved incongruently to the participant’s assessment of the turn situation compared to when it behaved congruently. There were no significant effects in N1 (142 to 182 ms), P2 (177 to 237 ms), nor P3 (439 to 689 ms). This suggests that, in human-AV interaction interaction, ERP-based devices might, in the future, be able to identify critical situations. However, further research is needed to bring current findings from fundamental research closer to application.