Measurement of Distinguishing Features of Stable Cognitive and Physical Health Older Drivers

This paper presents the analysis of all driving during 1 year by 14 older drivers from the Candrive study. Data analytic techniques have been applied to this unique big data set that includes global positioning system and geographic information system data with an average of 1567 trips and 20 200 km for each driver. The drivers have stable general, cognitive, and physical health over the year. This paper identifies trip features and tests their potential to be used to distinguish between driver pairs and shows that trip length and duration allow 90.1% and 90.1% of the 91 driver pairs (14 drivers) to be distinguished (p <; 5 %). Driver velocity histograms are also shown to allow driver pairs to be distinguished. This paper explores the deceleration habits of the drivers by locating all deceleration events with a net velocity drop of >=4 km/h (average of 30 156 events per driver) and finds that the event mean and minimum deceleration values have different two-phase relationships for each driver. Trips taken by a driver are compared with the drivers' two-phase relationships and this paper shows that the measures achieve their maximum performance for trips with >=8 events. The minimum measure is better (correlation of 0.74) than the mean measure (0.54). These relationships provide a measure of a driver's unique behaviors and could allow more personalized autonomous driving systems and capture features of the deceleration habits for the driver that can be used to distinguish between drivers.