Exploring the Relationship between False Alarms and Driver Acceptance of a Pedestrian Alert System during Simulated Driving

Abstract

In-vehicle pedestrian-alert-systems (PASs) can be prone to ‘false positive’ declarations, with the likelihood of false interventions increasing as time-to-collision (TTC) extends. A high number of false alarms can annoy drivers and lead to poor acceptance and low trust in the technology. To explore this relationship, 24 experienced drivers negotiated a single-lane urban high-street – moderately populated with pedestrians – during 12 five-minute drives in a medium-fidelity driving simulator. PAS warnings were presented in response to pedestrians who approached the roadside, either as a static visual alert icon presented on a HUD and/or auditory icon. The number of accurately detected pedestrians (i.e. those who entered the roadway rather than waiting at the kerbside) decreased with increasing TTC, giving rise to ‘false positive alarms’. Subjectively, participants associated the highest level of trust, confidence and desirability, and lower levels of annoyance, when warnings were presented at intermediate TTCs (3.0 and 4.0-seconds, corresponding to false-alarm rates of 40% and 60%, respectively); trust and confidence reduced significantly with both increasing and decreasing TTC. Driving performance data show that earlier warnings encouraged drivers to begin braking sooner and apply braking force more gradually, ultimately stopping further from the pedestrian – on average 18.0m following 5.0-second warnings compared to 6.2m with 2.0-second warnings. Nevertheless, evidence suggests that some drivers may have disregarded the system at longer TTCs, choosing to rely on their own judgement. The results have implications for the design, evaluation and acceptance of PASs.