This paper discusses a research method used to evaluate the design of the alerting logic of automotive active safety systems and presents an example of how this method can be used to support the design of pedestrian collision warning alerts. Three questions therefore arise: First, how can we collect a measure of the acceptability of an alert to a wide variety of situations? Second, how consistent is that measure across contrasting samples of drivers? Finally, how to use the measure in designing alerting logic?
We describe an empirical approach to quantifying the relative level with which drivers are likely to accept pedestrian alerts by a night vision system. The study had two parts: a field operational test (FOT) that gathered a set of 302 video clips of pedestrian alerts with a night-vision system, and a post-hoc or retrospective ratings experiment in which volunteers viewed the clips and rated the relative acceptability of the alerts. We document the consistency of these subjective ratings across groups of raters with different levels of experience with the system. This finding supports the argument that laboratory reviews of FOT data are likely to generalize across the population of drivers.
The derived measure of acceptance was then used to investigate a range of contextual and quantitative factors likely to influence driver acceptance of alerts to pedestrians issued by a night vision active safety system. Least squares regression revealed that nominal characterization of pedestrian location and motion and two quantitative measures – minimum separation and time to closest approach - explain almost 70% of the variance in driver ratings and do not interact. We discuss the implications of this finding for the specification of the system’s alerting strategies.