| Sumario: | Purpose As the portion of the aged population grows rapidly in South Korea, traffic injuries that are caused by older drivers have emerged as a public concern because the number of traffic crashes and injuries has increased. Older drivers have, in general, a degraded driving performance that reduces their ability to avoid risks (McGwin et al., 2000). As a result, their driving can be characterized as slow driving speed and wide deceleration and acceleration range (Horberry et al., 2006). However, the driving characteristics of older drivers were not fully unveiled due to data limitations. The objective of this study is to use large-scale actual driving records of older drivers to characterize the driving characteristics of older drivers and compare them with other drivers. Method The large-scale driving records of taxi drivers are collected using a digital tachograph (DTG), a naturalistic driving data collection device installed on commercial vehicles. Since DTG records vehicle speed, acceleration, steering angle, and location information every second, it is possible to analyze the driving patterns of drivers. In particular, these driving patterns are multidimensional data with non-linear characteristics (Chen et al., 2019). Therefore, to reflect on the attributes of the data and analyze the driving characteristics of drivers, a deep learning-based analysis methodology needs to be used. In this study, the normal driving patterns of taxi drivers are categorized using convolutional autoencoder-based deep clustering. In addition, an abnormal driving score is derived for each driving behavior, and driving characteristics according to drivers' age are analyzed. Results and Discussion Driving behaviors were characterized by different age groups. For the drivers whose age is 60 or older, the portion of abnormality - deviation from the normal driving behavior - increased in their acceleration and deceleration. It means that older drivers tend to be more frequent and have harsh pedal operations due to cognitive decline. On the other hand, the abnormality diminishes for drivers of 70 years or older in right-turning cases. The results of this study confirmed that the driving behavior varies according to the driver's age. Therefore, when establishing safety driving education programs and policies for older drivers, it is necessary to consider the differences in driving characteristics according to age.
|
|---|