Derivation of riding risk precursors using 100 delivery motor scooter naturalistic riding study

Abstract

The rapid growth of the delivery service market in Korea due to the impact of COVID-19 has resulted in an increase in crashes associated with delivery motor scooters. In particular, required minimum delivery time, which is an important factor for food delivery service, can lead to hazardous riding situations leading to traffic crashes. Although the food delivery service industry is continuously increasing, effective measures to improve the traffic safety of delivery motor scooters are insufficient. This study derived precursors in order to detect risky riding events using real-world naturalistic riding study data. It is essential to understand the riding characteristics of food delivery motor scooters to conduct the riding safety monitoring in more scientific and automated manners. Various candidate precursors were derived from riding characteristics data collected from GPS sensors and inertial measurement unit sensors. A decision tree model was then adopted to classify unsafe and normal riding events in order to determine the priority of precursors. A classification accuracy of 95.7% was obtained using three salient riding risk precursors including the norm of the angular velocity, which represents composite vector quantity of 3-axis measurements, acceleration, and X-axis angular velocity. The results of this study are expected to be used as a fundamental data to prepare for riding safety management systems that contribute to enhancing the safety of food delivery motor scooters.