Development of Aggressive Driving Detection Method for two-wheeled vehicles

Abstract

In the rapidly growing two-wheeled vehicle industry, there is an increasing need to detect and prevent aggressive driving behaviors. In this study, we propose an effective method to identify aggressive driving behaviors of simulated two-wheeled vehicles using limited data from the Carla simulator. We focus on two types of aggressive driving behaviors: sharp turning and sharp lane changing. By utilizing acceleration, angular velocity, and position data, we establish a foundation for detecting these behaviors. For sharp turning, a multi-stage classification method is adopted using gyroscope and acceleration data. Sharp lane changings are detected using an XGBoost mode