Warning reliability and driver performance in naturalistic driving
Authors: John M. Sullivan, Omer Tsimhoni, Scott E. Bogard.
This study examines how naturalistic driving performance is influenced by the perceived reliability of an in-vehicle warning system using a unique measure of perceived reliability. Background: Prior studies of warning reliability conducted in simulator and test-track experiments demonstrate that the objective reliability of a warning can influence a driver's responsiveness to that warning. Methods: Driver responses to lateral drift warnings (LDWs) were examined to determine if the latency to initiate a corrective steering response was related to the subjective reliability of prior system performance. An operational definition of subjective reliability was developed based on measures of the proportion of LDWs responded to by a steering correction in the preceding 24-hr period - the day proportion. Age, gender, weather condition, light level, road class, weekday status, and day proportion were used to model correction latency in a linear model. Results: A main effect of day proportion was found, suggesting that reaction time to respond decreases by about 375 ms as the day proportion increases from 0 to 1. Conclusion: The study illustrates a method of quantifying subjective reliability and performance using naturalistic data. The results suggest that latency to make a steering correction is inversely related to the perceived reliability of the warning system in the 24 hr preceding the warning. Application: The results have direct application to the method of assessing the efficacy of in-vehicle collision warning systems, suggesting that use of a perceived reliability measure may better predict a driver's disposition to respond to a warning.