Two-lane two-way overtaking decision model with driving style awareness based on a game-theoretic framework

Abstract

Overtaking on two-lane two-way (TLTW) highways is often associated with a high risk of crashing. However, existing models of TLTW overtaking decision, either mechanism- or learning-based, cannot handle well the dynamic coupling among the interacting drivers. For accurate overtaking modelling, it is crucial to consider the uncertainties of interacting vehicle behaviours, especially their driving styles. To address these needs, we propose a TLTW overtaking decision model using a level-k game theoretic framework, which can consider the mutual influences between the ego and oncoming vehicles of different driving styles. A dataset is built based on the TLTW overtaking experiments with two instrumented vehicles, then PCA and k-means clustering are used to classify three driving styles, i.e. aggressive, normal and conservative. By comparing the model predictions with the experiment data, the statistics and case studies show that the proposed model with driving style awareness can accurately describe driver decisions in TLTW overtaking.