Individual Stable Driving Pattern Analysis for Evaluating Driver Readiness at Autonomous Driving Levels 2 and 3

Autonomous vehicles have recently attracted considerable attention as a new type of transportation system that can improve convenience and quality of life. However, current versions of autonomous vehicles still require human intervention in specific situations, and in certain situations the driving mode must be switched from autonomous to manual. It is difficult to determine whether driver readiness is perfect at the moment at which the driving mode is switched because of a lack of concentration and tension. A certain amount of time is required to recover a stable driving state after taking over the controls. A stable driving pattern differs by driver, so it is important to identify and classify individual patterns to determine the time of take-over request (TOR). In this study, we attempt to measure the similarities among individual driving patterns and propose a classification model for the stable driving patterns of individuals. We first applied hierarchical clustering based on dynamic time warping and bag-of-patterns to confirm that the driver has his or her own driving pattern. We then improve the accuracy of driver classification for each driving event and implement a real-time application using the long-short-term memory methodology. We show that the classifier has relatively good performance for driving event section data, particularly for sudden stops and congestion events.